US 5,072,291 AGrant
Image Information Signal Processing Apparatus for Improving Reproduced Image Quality by Discriminating the Type of Input Image and Selecting a Particular Processing in Accordance Therewith
Issue Date:1991-12-10
•10 Claims
•11 Drawing Sheets
Abstract
In an image display system, a picture quality improving circuit is provided for comparing a distribution, in a pixel array of a specified size, of image characteristic signals derived from image information signals each indicating one pixel of the input image with a previously examined distribution of image characteristic signals corresponding to each type of image, to thereby discriminate the type of input image. Further, a binary encoding/dithering selecting circuit responsive to the decision made on the picture type by the image quality improving circuit is provided to binary encode the image information signal applied to the image information output device on the basis of a fixed threshold value, or to dither the image information signal on the basis of variable threshold values, in accordance with the discriminated type of input image. Image information signals corresponding to a character/line image are binary encoded by the fixed threshold value, and the image information signals corresponding to a halftone image or a continuous-tone picture are dithered.
Metadata
Assignee
- Kabushiki Kaisha Toshiba
Inventor
- Hidekazu Sekizawa
Application Information
Application Number:US 6360076
Filing Date:1991-01-03
Priority Date:1984-03-29
Art Unit:268
Classifications
IPC:
H04N 146
Field of Search:
35838275;456;455;466;467;46252;54
Patent Drawings (11 sheets)
Description
Background of the Invention
The present invention relates to an image information processing apparatus and, more particularly, to an image information signal processing apparatus which is used for displaying, with high quality, different types of picture images read by an image sensor from a document.
Recently, image processing systems have been strongly required to deal with a variety of types of office documents, which include color graphics images in addition to normal character/line-based documents. Such graphics images may contain continuous-tone picture images, such as photographs, and pseudo continuous-tone picture images, such as newspaper dot-photographs that are constructed by arranging various sizes of dots of three primary colors so as to provide a coarse photographic image which tries to "stimulate" the original natural photograph.
The "bi-level" images may be defined as images of two distinct density levels, which typically include the character/line-based document images. The pseudo continuous-tone picture images may also be categorized in the bi-level images. These pseudo continuous-tone picture images are sometimes called "dot-pattern" picture images, "dot photograph" picture images, "halftone" images, or "digitized" photograph images.
With presently available image output devices such as liquid-crystal display units or thermal printers, bi-level images can be displayed very stably. Where a general image which contains a continuous-tone picture image such as a photograph image is displayed or printed by these output devices, a continuous-tone image display/print scheme using a dithering technique is conventionally used, wherein a threshold value used for bi-level quantization is varied in accordance with a random or pseudo random function during the continuous-tone picture image information is being quantized into bi-level information.
With such a technique, however, when an image is displayed or printed with a number of gradation, not only does the resolution decrease but irregular noise is liable to be produced. This results in the image quality of an input image containing various types of image components being degraded.
An apparatus is known, which is arranged to input an image using a color image sensor equipped with spatially-divided color filters and output an output signal from the color image sensor to a color printer to reproduce the image. When an input document image containing a color continuous-tone picture image is copied by such an apparatus, dots forming the continuous-tone picture image are reproduced with increased noises, which results in that the colors in the printed document cannot be reproduced precisely.
An improved image information processing system has been disclosed in a copending U.S. patent application Ser. No. 556,323 filed Nov. 30, 1983 (U.S. Pat. No. 4,663,662 issued May 5, 1987); entitled "Picture Signal Processing System Suitable For Displaying Continuous Tone Pictures" and assigned to the same assignee as this patent application. According to one of the methods which are disclosed in this application, the local density variation of an input image is detected. If the density variation is greater than a predetermined specific value, an image information processing circuit quantizes an input image signal into a bi-level image signal on the basis of a fixed threshold value. On the other hand, if the local density variation is smaller than the specific value, the circuit dithers the input image signal.
According to the other method of the patent application, a brightness signal and color information signals are averaged in order to display a color pseudo continuous-tone picture image being constituted by various sizes of dots in three primary colors. By averaging the color information signals, dot-noises can be eliminated. The local density variation of the input image is detected on the basis of the brightness signal that has the dot-noises eliminated therefrom. If this density variation is greater than a preselected specific value, the color information signals are subjected to bi-level quantization using a fixed threshold value. If the density variation is smaller than the specific value, the color information signals are subjected to the dither-processing.
Unfortunately, both of the above methods fail to provide successful reproduction of various types of image images, since a high grade of technical difficulty has been laid in precise detection of pseudo continuous-tone picture regions from within the input image, as well as in selection of the most suitable type of image processing for the detected image regions.
Summary of the Invention
An object of the present invention is to provide an improved image information signal processing apparatus for use in displaying different types of document images.
Another object of the present invention is to provide an improved image information signal processing apparatus for processing in a different manner according to different types of document images including a character/line-based image, continuous-tone picture images, such as photographs, and pseudo continuous-tone picture images known as "dot photograph" image to properly display these image components.
The image information signal processing apparatus of the present invention is used in an image display system including an image information input means for obtaining an image information signal including an input image by optically scanning the input image, and image information output means responsive to the image information signal to reproduce an output image corresponding to the input image. To enable the image information output means to pleasantly reproduce the output image, the image information signal processing apparatus comprises: image-type discriminating means for discriminating the type of the input image on the basis of a distribution, within a pixel array of a specified size, of image characteristic signals derived from image information signals each representing one pixel of the input image, wherein the image characteristic signals are each a binary Laplacian signal of an image information signal corresponding to one pixel of the input image; and means responsive to the image-type discriminating means to select one of processing for binary encoding the image information signal in accordance with the type of input image, the processed image information signal being applied to the image information output means.
Other and further objects of this invention will become obvious upon an understanding of the illustrative embodiments to be described or will be indicated in the appended claims, and various advantages not referred to herein will occur to one skilled in the art upon employment of the invention in practice. has been decided by the picture type discriminating means to be the third type i.e., continuous-tone picture.
Brief Description of the Drawings
FIG. 1 shows a construction of a color copier which uses a image information processing system of the present invention;
FIG. 2 schematically shows an arrangement of a image quality improving circuitry of FIG. 1;
FIG. 3 shows in detail an arrangement of the quality improving circuit of FIG. 2 adapted for a brightness signal;
FIG. 4 is a view for explaining a pixel array which is used to decide the type of a image;
FIGS. 5A, 5B and 5C show image information signals which correspond to a halftone, character image and continuous-tone image, respectively;
FIGS. 6A and 6B show frequencies of occurrence of bi-level patterns of Laplacian signals for a character image and halftone images such as a dot photograph image, respectively;
FIGS. 7A, 7B and 7C show bi-level patterns of Laplacian signals for character images, which occur with high frequencies;
FIGS. 8A, 8B and 8C show bi-level patterns of Laplacian signals for halftone images, which occur with high frequencies;
FIG. 9 is a view for explaining a decision ROM shown in FIG. 3;
FIG. 10 shows a geometric arrangement of pixels for explaining decision operations performed in the decision ROM;
FIGS. 11 to 13 show various bi-level patterns of image signals, adapted for explaining decision codes stored in the decision ROM;
FIG. 14 shows a image quality improving circuit for a color difference signal of FIG. 2;
FIG. 15 shows a second embodiment of the present invention;
FIGS. 16A to 16C, 17A to 17C, 18A to 18D, 19, and 20A to 20D are views for explaining a image-type deciding operation in the second embodiment of the present invention;
FIG. 21 shows the construction of a color copier which uses the second embodiment;
FIG. 22 shows a third embodiment of the present invention;
FIG. 23 shows an image type discriminator which is preferably used in the embodiments of the present invention;
FIGS. 24A and 24B show typical bi-level patterns of Laplacian signals for a character/line-based image;
FIGS. 25A and 25B show typical bi-level patterns of Laplacian signals for a halftone image such as a dot-picture;
FIGS. 26A and 26B show typical bi-level patterns of Laplacian signals for a continuous-tone image such as a photograph;
FIG. 27 shows apart of an image type discrimination table prestored in a ROM included in the circuit shown in FIG. 23;
FIG. 28 shows a fourth embodiment of the present invention; and
FIG. 29 shows the internal circuit arrangement of an image quality improving circuit provided in the embodiment shown in FIG. 28.
Detailed Description of the Preferred Embodiments
A image information processing apparatus according to an embodiment of the present invention, as shown in FIG. 1, is arranged to suitably process a image information signal to be applied to a image reproduction device by discriminating the characteristic of picture within a local area of a document, in accordance with distribution of rates of variation of the image information signal, corresponding to that local area of the document, read by means of an image sensor. The image information processing apparatus of this embodiment uses a scanning-type proximity sensor as its image input section, and a thermal printer as its image-displaying output section.
Referring now to FIG. 1, a proximity sensor 11 reads an input document 2 carrying a color image thereon, while scanning the document 2 in a direction indicated by an arrow, to produce a image information signal. In the proximity sensor 11, a linear light source 1 irradiates the document 2 and light reflected therefrom is focused on a linear CCD image sensor 4 having CCD chips through a self-focusing rod-lens array 3 which is known under the trade name of "SELFOC". On each sensor element of the image sensor 4 one of color filters for red (R), green (G) and blue (B) components is provided whereby electric signals of R, G and B components for one pixel are obtained. The image information signal contains signals indicating the R, G and B components for each pixel, which are serially read out from the image sensor 4 for each pixel. In this embodiment, one pixel has a length of 84 .mu.m and, within this one pixel, three sensor elements are contained.
The image information signal which has been read out from the image sensor 4 is amplified in an amplifier 5 and then is converted, by an analog-to-digital (A/D) converter 6, into a digital signal of 8 bits. This digital signal is applied to a normalizing circuit 7, which is intended to compensate for variation in sensitivity of the CCD sensor elements of the image sensor 4 and variation in optical characteristics of the color filters, eliminate fixed noise in the CCD, and compensate for variation in the amount of light emanated from the light source and unwanted distribution of the amount of light.
The signal thus normalized is applied to a matrix circuit 8 to be converted from R, G and B signals into a brightness signal I and color difference signals C1 and C2. These signals are applied to a image quality improving circuit 10 as described later which constitutes a main part of the image information processing apparatus. The picture quality improving circuit 10 is arranged to decide the type of the input image and properly processes the brightness signal I and color difference signals C1 and C2 in accordance with the input document. After having been processed in this way, the brightness signal I and color difference signals C1 and C2 are applied to a color conversion circuit 12. The signal system prior to the color conversion circuit 12 handles the additive primaries of R, G and B. The color conversion circuit 12 converts these additive primary signals into subtractive primary signals of magenta (M), cyan (C) and yellow (Y) suitable for the subsequent print in a color printer 16.
The M, C, Y color signals which have been output from the color conversion circuit 12 are applied to a binarization/dithering selecting circuit 14, in which they are binarized or dithered in accordance with a decision made by the image quality improving circuit 10 on the input image (a continuous-tone image such as a photograph, a character/line-based image such as a character document image, or a halftone image such as a pseudo-photographic image called the "dot photograph" image). Thereafter, they are applied to the color printer 16, which reproduces a pleasant color output image corresponding to the color input image by virtue of the image quality improving circuit 10 and the binary encoding/dithering selecting circuit 14.
The image quality improving circuit 10 will now be described in detail. It includes, as shown in FIG. 2, an image quality improving circuit 19a for processing the brightness signal I, and image quality improving circuits 19b and 19c for processing the first and second color difference signals C1 and C2, respectively. The brightness signal and first and second color difference signals which have respectively been processed by those image quality improving circuits are denoted by I, C1 and C2, respectively.
The brightness-signal image quality improving circuit 19a is shown in FIG. 3. This circuit includes a Laplacian calculating circuit (hereinafter, referred to as "a first circuit") 20, which receives the brightness signal I and calculates its Laplacian with respect to each pixel to produce a Laplacian signal. The Laplacian signal output from the first circuit 20 is applied to a second circuit 30. In this second circuit 30, the Laplacian signal is binary encoded on the basis of a fixed threshold value. Further, in this circuit 30, a pattern of the binary Laplacian signals corresponding to a pixel array of a specified size is compared with predetermined patterns of binary Laplacian signals in a pixel array of the same size. These patterns are previously prepared in the circuit 30 in accordance with the characteristics of various images. The result of comparison is represented in the form of a two-bit configuration. The type of input image is roughly decided by way of this pattern comparison.
The Laplacian signal from the first circuit 20 is also applied to a third circuit 40, which codes an absolute value of the Laplacian signal into a four-bit signal and adds together these coded signals corresponding to a specified number of pixels. The third circuit 40 is intended to examine the significance of the rough decision made by the second circuit 30 in accordance with the magnitude of the Laplacian signal. Therefore, for example, even when the second circuit 30 decides an input document by noise (in this case, the Laplacian signal is relatively small) resulting from stains on the input document, the third circuit 40 can inform the image information processing circuit of the fact that the decision made by the second circuit 30 is based on noise. The output of the third circuit 40 consists of upper four bits of the result of addition.
The outputs of the second and third circuits 30 and 40 are coupled to a fourth circuit 50 which is arranged to integrally decide the type of the input picture. The fourth circuit 50 decides whether the input image is a continuous tone picture, a line image or a halftone image. In order that the circuit 50 may precisely decide a portion of the input document, the result of a preceding decision on another portion of the input document near that portion being decided is fed from a fifth circuit 60 back to the fourth circuit 50. That is, output signals of the second, third and fifth circuits are utilized as image-type decision input signals of the fourth circuit 50. The output signal J of the image-type decision circuit 50 is composed of two bits, the most significant bit of which is applied to the fifth circuit 60. The output signal of the fifth circuit 60 is composed of four bits.
The two-bit output signal J of the image-type decision circuit 50 is applied to a sixth circuit 70 for processing the picture information signal. The image information signal processing circuit 70 receives the brightness signal I to compensate for its high frequency components and average the brightness signal I with respect to a specified number of pixels. Further, the circuit 70 is responsive to the output signal J of the image-type decision circuit 50 to selectively output the high-frequency compensated signal, the averaged signal, or the brightness signal.
In more detail, the Laplacian calculating circuit 20 includes a delay circuit 21 for delaying the input brightness signal I, a low-pass filter (LPF) 23 for passing only low-frequency components of the input brightness signal I, and a subtractor 25 for calculating the difference (Laplacian) between the output signals of the delay circuit 21 and LPF 23. The calculation of Laplacian in accordance with this embodiment corresponds to a process of extracting the high-frequency components from the input signal. It is well known that both of them are qualitatively in coincidence with each other. LPF 23 may actually be arranged to average the picture information signal with respect to a specified number of pixels, e.g., 3.times.3 pixels. Constructing LPF 23 in a digital manner is easy. The output signal of the subtractor 25 is applied, as a Laplacian signal, to the second and third circuits 30 and 40.
The algorithm of the Laplacian operation used in this embodiment is such that, as shown in FIG. 4, the difference between the value of a center pixel i22 in the array of 3.times.3 pixels and a mean value of pixels ilm (l, m=1, 2, and 3) including i22 is calculated to obtain the Laplacian of the pixel i22. More specifically, the LPF 23 may comprise an adder circuit for adding together values of corresponding consecutive three pixels ilm on three consecutive scanning lines, and a multiplier circuit (which can be inexpensively constructed using ROM) for reducing an output value of the adder circuit to 1/9. The subtractor 25 calculates the difference between the value of the pixel i22 and a mean value of the 3.times.3 pixels ilm to thereby obtain a Laplacian of the pixel i22. In this way, the Laplacian of each pixel is obtained. The Laplacian signal is composed of 8 bits including one sign bit.
The eight-bit Laplacian signal from the first circuit 20 is binary encoded, in the second circuit 30, into "1" or "0" on the basis of a fixed threshold value by a binary encoder 31, which can be constructed of ROM. The binary Laplacian signal from the binary encoder 31, which corresponds to each pixel, is sequentially stored, for each scanning line, in line memories 32, 33, and 34 through a multiplexer 35. Three bits are read out from each of the line memories 32, 33 and 34 and a resultant nine-bit signal is applied to a ROM 36 as an address signal. The ROM 36 previously stores code information, which identifies the types of input images, resulting from distributions of binary Laplacian signals in the 3.times.3 pixel array which are examined beforehand with respect to continuous tone images, line images and halftone images. For this reason, the ROM 36 is accessed by an address signal from the line memories 32, 33 and 34, which indicates the distribution of binary Laplacian signals, to thereby provide a two-bit code information signal indicating the type of corresponding image. As stated before, the object of second circuit 30 resides in the rough decision of the type of image or the document.
The consideration on which the image information processing of the present invention is based will now be described.
FIG. 5A shows a brightness signal which is obtained by one-dimensional scanning of a halftone image by the proximity sensor 11 while, on the other hand, FIG. 5B shows a brightness signal which is obtained by one-dimensional scanning of a character picture by the sensor 11. As seen in FIGS. 5A and 5B, in the halftone image and character picture, great local variation occurs in the amplitude of the brightness signal which indicates the density of the picture. For this reason, it is difficult to discriminate between the dot and character images on the basis of such local variation in amplitude of the Laplacian signals. FIG. 5C shows a brightness signal which is obtained by one-dimensional scanning of a continuous-tone image such as a photograph. In this case, as can be seen, the Laplacian signal is small. Therefore, such a continuous-tone image can be sufficiently discriminated from the halftone image and character picture on the basis of the magnitude of the Laplacian signals. In order to discriminate between the halftone image and the character image, it is sufficient to examine the distribution of binary Laplacian signals corresponding to pixels in a specified pixel array.
Examination was made of the distribution of binary Laplacian signals with respect to each of a character picture and halftone image, the results were obtained as shown in FIGS. 6A and 6B. In the Figures, the abscissa represents the arrangement of binary Laplacian signals in the array of 3.times.3 pixels and the ordinate represents the frequency with which the arrangement of binary Laplacian signals occurs. The arrangement of binary Laplacian signals consists of a 9-bit data whose MSB and LSB correspond to ill and i33 in the 3.times.3 pixel array shown in FIG. 4, respectively. For example, where only the binary Laplacian signal corresponding to the pixel ill is 1, the 9-bit data is given by "100000000". At the left end of the abscissa, all binary Laplacian signals for the 3.times.3 pixel array are 0 and, at the right end thereof, all binary Laplacian signals are 1. As seen from the result shown in FIG. 6A, in case of character images, such patterns of binary Laplacian signals as shown in FIGS. 7A, 7B and 7C appear with a higher frequency. In this case, there is a tendency that not only white pixels but also black pixels are arranged side by side to some extent. This tendency coincides with the qualitative tendency of a character pattern.
On the other hand, the result shown in FIG. 6B is obtained with respect to halftone images. As seen, such patterns of binary Laplacian signals as shown in FIGS. 8A, 8B and 8C appear with a higher frequency. This indicates the qualitative tendency of halftone images that neither white pixels nor black pixels are continuously arranged but they have a periodic pattern. In FIGS. 7 and 8, portions where oblique lines are drawn each represent the binary Laplacian signal of "1" and portions where no lines are drawn each represent the binary Laplacian signal of "0".
Accordingly, the discrimination of the type of input images becomes possible by examining the magnitude of Laplacian signals and the distribution of binary Laplacian signals. For this reason, in the second circuit 30, the Laplacian signals are binary encoded and the address signal indicating the distribution of binary Laplacian signals in the 3.times.3 pixel array is applied to the ROM 36. In accordance with data shown in FIGS. 6A and 6B, ROM 36 stores code information adapted for discriminating the type of input images, in its storage location accessed by an address signal indicating an arrangement of binary Laplacian signals. In order to discriminate three types of images, i.e., halftone image, line image, and continuous tone image, ROM 36 stores code information composed of two bits (for example, 00, 01, or 11). For simplification, only the sign bit of Laplacian signal may be handled in place of binary encoding the Laplacian signal.
The Laplacian signal is also applied to the third circuit 40, in which the Laplacian signal is converted into an absolute value and also is subjected to non-linear conversion. ROM 41 converts the Laplacian signal, in a non-linear manner, into a 4-bit signal for the convenience of the subsequent additive operation, whereby Laplacian signals having opposite signs and an equal magnitude are converted into the same digital information of 4 bits. The 4-bit information from the ROM 41 is applied to a latch circuit 43 through a delay circuit 42 arranged to provide a time delay corresponding to the time delay caused by the second circuit 30. The latch circuit 43 stores information signals corresponding to three pixels. The three-pixel information signals are added together by adders 44 and 45. The most significant 4 bits of the 5-bit sum is applied to the image-type decision circuit 50. The third circuit 40 examines the magnitude of Laplacian signals to cooperate the fourth circuit 50 for deciding the type of input images. Through a simple image-type decision by the second circuit 30 based on the comparison between distribution patterns of binary Laplacian signals, stains or the like of the input document can be determined to be one of halftone image, continuous tone image, and line image. In case of stains, the Laplacian signals are relatively small. For this reason, through examining the magnitude of the Laplacian signals, the image-type decision circuit 50 is prevented from making an erroneous decision such as stains being one of the above-mentioned three types of images, so a more precise decision of the picture type becomes possible.
The image-type decision circuit 50 is formed of ROM. Following address input signals are applied to the ROM as shown in FIG. 9. That is, a 2-bit signal from the ROM 36 are applied to inputs 1 and 2 corresponding to two most significant bits, a 4-bit signal from the third circuit 40 are applied to inputs 6 to 6, a 3-bit signal from a one-line delay circuit 62 as later described are applied to inputs 7 to 9, and a 1-bit signal from a one-pixel delay circuit 61 is applied to input .circle. 10 corresponding to the least significant bit. As later described, the ROM 50 evaluates a minute picture area (4.times.5 pixels) on the basis of address signals applied thereto.
The ROM 50 has two-bit outputs .circle. 11 and .circle. 12 . That is, the output signal J of the ROM 50 is 11, 10, 01 or 00. The meanings of these output signals are as shown in Table 1 below.
The output signal "00" of the total image type decision ROM 50 corresponds to a halftone picture, indicating that the picture information signal is to be averaged. The output signal "01" corresponds to an ordinary image which falls outside the category of halftone image or character/line images. The output signal "10" corresponds to a image having sharp edges, indicating that the high frequency range of the image information signal has to be stressed. The output signal "11" corresponds to a character/line image, indicating that the high frequency range of the image information signal has to be stressed. That the most significant bit output .circle. 11 of decision ROM is 1 indicates that the variation in density of the image read is great. The output signal of the decision ROM is applied to binary encoding/dithering selecting circuit 14 as later described. When the output signal J of the image-type decision ROM is 11, the binary encoding/dithering selecting circuit 14 selects the binary encoding of the image information signal based on a fixed threshold value. On the other hand, when the output signal J of the decision ROM is 10, 01 or 00, the circuit 14 selects the dithering of the image information signal.
MSB .circle. 11 of the output signal J of decision ROM 50 is applied to the fifth circuit 60 as well. This circuit 60 includes one-pixel delay circuit 61 and one-line delay circuit 62. The output bit delayed by a time corresponding to one pixel by the one-pixel delay circuit 61, and the oldest three bits of the output bits of decision ROM delayed a time corresponding to one line by the one-line delay circuit 62 are applied to the ROM 50 as mentioned before.
The geometric arrangement of pixels corresponding to the bit signals applied from the fifth circuit 60 to the image-type decision ROM 50 will now be described. Referring to FIG. 10, it is assumed that a pixel which is now being processed is i0,0. Thus, the output signal of the one-pixel delay circuit 61 is an MSB output of the decision ROM 50 for the pixel i-1,0 adjacent to the pixel i0,0. The three bit outputs of the one-line delay circuit 62 represent outputs of ROM 50 for pixels i-1,-1; i0,-1; and il,-1 on the immediately preceding line. That is, one-line preceding decision outputs and one-pixel preceding decision output are fed back to decision ROM 50. The aim of this feedback is to reduce decision error possibly made in evaluating a pixel now being processed, by taking account of decision results for surrounding pixels. For example, if the surrounding pixels have been decided to correspond to part of a character, the pixel which is now being processed can also be expected to correspond to part of the character. For this reason, account is taken of the previous decisions for preceding pixels.
The contents stored in the decision ROM 50 will now be described. As shown in FIG. 11, with respect to a case where all the MSBs of output signals J of decision ROM for pixels i-1,-1; i0,-1; il,-1; and i-1,0 are 0, i.e., where a image indicates a slow variation in density, in the decision ROM are stored the contents as shown in Table 2.
In Table 2 above, the numeric values 0, 1, 2, and 3 stand for two-bit information 00, 01, 10, and 11, respectively.
The above Table indicates that when 1=2=0, and 3=4=5=6=1, the output J of decision ROM 50 is "00". On the other hand, when 1=2=1, and 3=4=5=6=1, the output J of the decision ROM is "11". To explain the significance of this Table qualitativly, the surrounding pixels have been determined such that the image density thereof slowly varies and thus, even if a density variation occurs with respect to that pixel now being processed it is decided that such variation results from noise and the picture density still varies slowly.
Next, as shown in FIG. 12, with respect to a case where i-l,-l=il,-l=0; and i0,-l=i-l,0=l, the contents as shown in Table 3 below are stored in the decision ROM.
Next, as shown in FIG. 13, in the case of i-l,-l=il,-l=i0,-l=i-l, o=l, the contents as shown in Table 4 below are stored in the decision ROM.
Each of these tables is provided in the decision ROM for every one of 16 (=2.sup.4) values which can be taken by the 4-bit output signal of the feedback circuit 60.
A sixth circuit 70 selectively applies one of the three signals--brightness image-information signal I, high-frequency stressed brightness signal, and averaged brightness signal--to the color conversion circuit 12 in accordance with the image quality, i.e., in accordance with the output signal J of the decision ROM 50. As shown in FIG. 3, the sixth circuit 70 includes delay circuits 71 and 72 provided for synchronizing the output of this circuit 70 with the output signal of the decision ROM 50, a subtractor 73 to prepare a differentiating signal, a multiplier (ROM) 74, an adder 75 to prepare a high-frequency stressed signal, and a selecting circuit 76. Output signals of the delay circuit 21 and LPF 23 of the first circuit 20 are delayed by the delay circuits 71 and 72, respectively. The output signals of the delay circuits 71 and 72 are applied to the subtractor 73 in which the differentiating signal is produced. This differentiating signal is multiplied by K in the multiplier 74. The output signal of the multiplier 74 and the output signal of the delay circuit 71 are added together in the adder 75, to form the image information signal whose high frequency components are stressed. The output signal of the delay circuit 71, i.e., the brightness signal I, the output signal of the adder 75, i.e., the high-frequency stressed brightness signal, and the output signal of the delay circuit 72, i.e., the brightness signal averaged with respect to 3.times.3 pixels are applied to the switching circuit 76. This switching circuit 76 selects one of those three image information signals in accordance with the 2-bit output signals J of the decision ROM 50 and applies it to the color conversion circuit 12.
In this way, the brightness signal I improved in accordance with the image quality is obtained from the image quality improving circuit 19a. This brightness signal I, the output signal C1 of the image quality improving circuit 19b for the first color difference signal, and the output signal C2 of the image quality improving circuit 19c for the second color difference signal are applied to the color conversion circuit 12, the output signals M, C, and Y of which are binary encoded or dithered in the binary encoding/dithering selecting circuit 14 responsive to the output signal J of the decision ROM 50.
The image quality improving circuits 19b and 19c for color difference signals, as shown in FIG. 14, can be constituted by LPFs 120a and 120b like LPF 23 shown in FIG. 3, respectively. Namely, the color difference signals C1 and C2 are averaged with respect to nine pixels. Only the averaging process is sufficient for the image quality improvement in the color difference signals. This is because the resolution of human vision for colors of a is inferior to that with respect to the brightness. For this reason, to sufficiently reproduce a, such a complicated processing as required of the brightness information is not necessary for color information.
The output signals I, C1 and C2 of the image quality improving circuit 10 are applied to the color conversion circuit 12 to be converted into M, C, Y signals suitable for printing. The color conversion circuit 12 is formed of a ROM which stores suitable conversion tables. The output signals M, C, and Y of the color conversion circuit 12 are subjected, in the binary encoded/dithering selecting circuit 14, to binary encoding based on the fixed threshold value or dithering processing.
The binary encoding and dithering processings will now be described. In principle, character/line images are binary encoded on the basis of a fixed threshold value, while other images are dithered. However, the character/line images must be displayed with high resolution. Accordingly, a image information signal whose high frequency components are stressed is necessary for reproducing the character/line images. On the other hand, halftone images are dithered since they must be displayed in continuous tones. In many cases, a halftone image information signal has very high frequency components, which is attributable to the nature of halftone images. For this reason, when the halftone image signal is dithered as it stands, noise will appear prominently. Accordingly, an average signal is used for the halftone image signal. A image information signal corresponding to a image other than character/line image and halftone image are used as it stands.
When the output signal J of the decision ROM 50 is "11" or "10" corresponding to character/line pictures, the switch circuit 76 selects the high-frequency stressed image information signal from the adder 75. When the output signal J is "00" corresponding to dot images, the switching circuit 76 selects the averaged image information signal from the delay circuit 72. In case of images other than character/line images, sharply edged images, or dot pictures, the output signal J is "01", causing the switching circuit 76 to select the output signal of the delay circuit 71, i.e., the non-processed brightness signal I.
As mentioned before, the image information signals I, C1 and C2 are converted, in the color conversion circuit 12, into color signals M, C and Y each of 6 bits. In this case, the color conversion circuit 12 may be arranged to produce a black-and-white signal to intensify contrast. These color signals are binary encoded or dithered, in the binary encoding/dithering selecting circuit 15, in accordance with the output signal J of the decision ROM 50 in the image quality improving circuit 10. The binary encoding/dithering selecting circuit 14 processes each color signal. As mentioned above, J=11 stands for character/line images, J=00, halftone images, and J=01 or 10, other pictures. However, the binary encoding/dithering selecting circuit 14 is arranged to select, when J=11, the binary encoding of image information based on a fixed threshold value, and, when J=00, 01 or 10, binary encoding based on the dithering technique. The binary encoding/dithering switching circuit 14 can be constructed in accordance with the teaching described in the above-mentioned prior application. Briefly speaking, the circuit 14 may be arranged such that a fixed threshold value memory for simple binary encoding, a threshold value memory for dithering binary encoding, and a comparison circuit are prepared and one of the memories is coupled to the comparison circuit in accordance with the output signal J of the decision ROM. Further, the binary encoding/dithering switching circuit 14 may be inexpensively constructed with the use of a ROM common to color signals and a multiplexing technique. The color printer 16 realizes color printing by thermal printing technique based on the plane sequence system using sequentially processed color signals.
According to the above-mentioned arrangement, in the case of character/line pictures, a image information signal read by the image sensor has its high frequency components intensified and is then binary encoded on the basis of a fixed threshold value. Thus, a clear output image can be obtained from the color printer. On the other hand, where image is a halftone image, the image information signal is averaged and then dithered. Therefore, the noise which has been produced in reading the halftone picture are eliminated. Where picture is other picture than character/line image and halftone image, the image information signal is processed by ordinary dithering technique, whereby a pleasant continuous tone image can be presented.
A image information processing apparatus according to a second embodiment of the present invention will now be described with reference to FIGS. 15 to 21. In FIG. 15, a image information signal I is applied to a simple binary encoding circuit 502 through an input terminal 501. The image information signal I is an 8-bit digital signal corresponding to an output signal delivered from the image sensor as in the first embodiment. To simply binary encode the signal I the most significant bit of signal I may be used. One-bit output signals of the binary encoding circuit 502 are sequentially stored, for each line, in one-line memories 504, 505, and 506 through a multiplexer 503. 3 bits are simultaneously read from each of the one-line memories 504, 505, and 506 to be applied to 9-bit address input of a first decision ROM 507. This ROM 507 stores first image-type decision codes each composed of 2 bits, e.g., "10" for a pattern (including part of a continuous-tone picture pattern) near to character/line image, "01" for a pattern near to halftone images, "11" for a pattern near to continuous-tone image, and "00" for other image patterns. The 2-bit output signal of the ROM 507 is applied to an integral decision ROM 508.
On the other hand, the image information signal I from the input terminal 501 is binary encoded in a comparator 509 on the basis of variable threshold values, i.e., dithered. This comparator is supplied with threshold value data from ROM 501 storing 3.times.3 threshold values. To the address input of ROM 510 are coupled a counter 511 for counting pixels in the main scanning direction (line direction) of the array of 3.times.3 pixels, and a counter 512 for counting pixels in the sub-scanning direction (the direction perpendicular to the line direction). Thus, addresses of ROM 510 are sequentially designated, whereby threshold value data arranged in accordance with 3.times.3 pixel array are sequentially read out. As a result, the image information signal I is dithered in the comparator 509. The dithered image information signal is sequentially stored for each line in one-line memories 514, 515, and 516 through a multiplexer 513 3 bits are simultaneously read from these memories to access a second decision ROM 517. In this ROM are stored second image-type decision codes each composed of 1 bit, e.g., "1" for a continuous-tone pattern which slowly varies, and "0" for other image patterns. The second image-type decision code signal from the ROM 517 is applied to the decision ROM 508.
The decision ROM 508 is accessed by the first image-type decision 2-bit code signal, the second image-type decision 1-bit code signal, and the 1-bit signal (which indicates a preceding decision result) of a latch circuit 518 which holds a 1-bit output signal of ROM 508, to thereby integrally determine the type of an input image. The decision ROM 508 discriminates character/line images from other images in response to a 4-bit address input signal mentioned. A 1-bit decision output signal of the ROM 508 is coupled to a multiplexer 523 to selectively couple one of output signals 520 and 521 of one-line memories 505 and 515 to output 522.
The image-type decision principle used in the embodiment of FIG. 15 will now be explained. FIGS. 16A, 16B and 16C show part of binary encoded patterns well produced when binary encoding a image information signal corresponding to character/line images. As can be seen, with character/line images white or black appears somewhat continuously. Part of a continuous-tone image, as well, represents such a pattern. On the other hand, halftone image in may cases present such patterns as shown in FIGS. 17A, 17B and 17C. A variable-density image well presents such patterns as shown in FIGS. 18A, 18B, 18C and 18D. Part of characters also presents such patterns as with the variable-density image. The use of binary encoding patterns for 3.times.3 pixels would make it possible to roughly discriminate between character/line image, halftone, and continuous-tone image.
Image discrimination by binary encoding patterns based on the use of variable threshold values will now be described. FIG. 19 shows an example of variable threshold value table. The numeric values of 1 to 9 represent nine levels of picture density. Where 3.times.3 pixels each having a density of level 5 are binary encoded on the basis of this variable threshold value table, the resultant binary encoding pattern becomes as shown in FIG. 20A. Where image density varies slowly as in a variable-density image, the density variation within the 3.times.3 pixel array is, usually, to an extent of .+-.1 or .+-.2 levels. For example, where the 3.times.3 pixels each have a density of level 5 and the variation in density is .+-.1, the binary encoding pattern varies as shown in FIG. 20B or 20C. With respect to such a pattern, the second decision ROM 517 stores "1" which indicates the continuous-tone image.
Where a density level variation within the 3.times.3 pixel array is as wide as 4 levels as shown in FIG. 20D, the image cannot be regarded as a variable-density image whose density slowly varies. With respect to such a pattern, the ROM 17 stores "0". In this way, a variable-density image can be discriminated by the second decision ROM 517 from other images.
Images are continuous. Therefore, the decision on the preceding pixel is very likely to apply to the pixel which is now being processed. Therefore, if a feedback loop having a memory such as latch circuit 518 is provided for decision ROM 508, the accuracy of the decision will be increased.
The embodiment of FIG. 15 may also be applied to processing of color information signals. As shown in FIG. 21, R, G, and B picture information signals are converted, in a matrix circuit 70, into a brightness signal I, and color difference signals C1 and C2. Further, color conversion circuit 572 provides color picture information signal M, C, and Y. These color picture information signals are applied to a color printer 574 through a binary encoding/dithering selecting circuit 573. The brightness signal I from matrix circuit 570 is also applied to a decision circuit 571 as shown in FIG. 21. Decision circuit 571 includes a total decision ROM, which switches the binary encoding/dithering selecting circuit 573. The character/line images are binary encoded on the basis of a fixed threshold value, while the other images are dithered. In this way, even in case of color images, character/line images can be displayed with high resolution, and continuous-tone images and halftone images can be faithfully reproduced in proper continuous tone.
A tone information processing apparatus according to a third embodiment of the present invention will now be described with reference to FIG. 22. The image information signal is applied to a binary encoding circuit 602 through an input terminal 601. In this binary encoding circuit 602, the image information signal is binary encoded on the basis of a fixed threshold value. As stated before, for simplification, the most significant bit of the image information signal may be used as the binary signal. The binary signal from the binary encoding circuit 602 is sequentially stored for each line in one-line memories 604, 605, and 606 through a multiplexer 603. 3 bits are simultaneously read out from the one-line memories to access ROM 607. This ROM 607 previously stores a code "1" with respect to a character pattern, and stores a code "0" with respect to other image patterns.
An output code of ROM 607 is applied to a decision ROM 608, which is accessed by 1-bit output signal of ROM 607 and a 1-bit output signal of a latch circuit 609 which holds an output signal of ROM 608. The decision ROM switches the binary encoding/dithering circuit 610 so as to binary encode, when the input image is a character, the image information signal on the basis of a fixed threshold values, and dither the image information signal when the input image is another image. The picture information signal thus binary encoded or dithered is applied to printer 611. In this third embodiment, by utilizing the continuity of characters in white or black as in the second embodiment, the characters are discriminated from other images. In case of the color images, the arrangement shown in FIG. 21 can be used.
Referring now to FIG. 23, another embodiment in accordance with the present invention is shown which includes an automatic input image type discrimination section. This image type discriminator may be also applied to any one of the aforementioned embodiments.
An input document image is optically read by image sensor 11 (see FIG. 1) to produce an electrical analog image information signal, which is then subjected to analog-to-digital conversion, normalization, compensation, and signal conversion, thereby to obtain a corresponding brightness signal I and color difference signals C1 and C2 in the same manner as described with reference to FIG. 1.
The signals I, C1, and C2 are supplied to the image quality improving circuit 10 shown in FIG. 1 including the Laplacian calculating circuit 20, which extract high-frequency components from the brightness signal I to generate a Laplacian signal binary-encoded on the basis of a fixed threshold value. The Laplacian signal may be 8-bit digital signal.
As shown in FIG. 23, a binary encoder circuit 800 called the "threshold value circuit" is provided to receive the Laplacian signal. This signal is binary-encoded by binary encoder 800 into "1" or "0" on the basis of a preselected threshold value. It is preferable that the binary processing should be performed with a threshold value slightly greater than 0 in light of noise, which reduces the noise influence.
The binary output signal of the circuit 800 is supplied to a delay matrix circuit 802, which is formed of three line delay memories 804a, 804b and 804c and a matrix of 1-pixel delay memories 806 which are arranged to provide a 4.times.4 image pattern as shown in FIG. 23. An output signal of the delay matrix circuit 802 representing the extracted image pattern is then supplied to a ROM 808.
The ROM 808 may correspond to the ROM 31 of the embodiment shown in FIG. 3. The ROM 808 previously stores necessary code information, which is used to identify the types of input images. While the previous ROM 36 is designed to deal with 3.times.3 image pattern, the ROM 808 can deal with more greater size of 4.times.4 pattern. With such an arrangement, the accuracy of image type judgment can be enhanced, although the total data capacity of the ROM 808 is 27 times greater.
The ROM 808 has therein an input/output conversion table wherein, when a certain data pattern is input thereto, a specific output data corresponding to the input data pattern is automatically generated as an "image type discrimination result." For example, when a certain type of 4.times.4 pixel matrix of local image pattern is input to the ROM 808, the corresponding output data is automatically searched inside the ROM table and extracted therefrom as an image type discrimination data. In this sense, it can be said that the ROM 808 has a function to automatically discriminate the property of an image based on an pattern of a binary high-frequency component extracted from the input image, whereby the input documents may be discriminated to be one of a character/line drawing image, a half-tone image such as a dot-photograph image, and a continuous-tone image such as a photograph. The designing of the ROM 808 will be described in detail later.
The 4-bit output data of the ROM 808 is then supplied, as an image type discrimination result, to a partial image type discrimination data processing circuit 810 for making, by arranging a plurality of pieces of image type data, a total image type judgment with respect to a greater region of the input image including several 4.times.4 pixel matrix sections.
The data processing circuit 810 includes a series array of 1-bit pixel delay memories 812a to 812g. The outputs of the first three delay memories 812a, 812b and 812c are connected in parallel to three inputs of 3-input/1-output switching circuit 814. The outputs of last three delay memories 812e, 812f and 812g are connected to another 3-input/1-output switching circuit 816 in such a manner that the output of the delay memory 812f is directly connected to a second input of the switch circuit 816 while the outputs of the remaining two delay memories 812e and 812g are insulatively "cross" each other and are connected to a third input and a first input of the switch circuit 816 as shown in FIG. 23. These switch circuits 814 and 816 operate in response to a switch control signal Sc which is externally supplied thereto. A decoder circuit 818 is provided which has two inputs to which the outputs of the two switch circuits 814 and 816 are supplied respectively.
The data processing circuit 810 is connected to a series circuit of 1-pixel delay memories 820 via an adder 822, with which a line delay memory 824 is associated. The outputs of series-connected delay memories 824 are connected in parallel to inputs of an adding circuit 826, which performs an averaging processing for the sequentially delayed pixel data with eight pixels being as a unit, thereby to produce a final image type discrimination data of the input image scanned by the image sensor 11.
The key to provide a high-accuracy image type discrimination in the ROM 808 is that the input/output data conversion table should be carefully prepared by an operator and stored in the ROM 808 as a "automatic image type discrimination table." The basic discrimination scheme of such ROM table may be designed on the basis of the static examination on the specific "nature" inherent to the three different types of input images.
More specifically, if the 4.times.4 pixel matrix of a local pattern is a part of the character/line drawing image, it is observed to provide a greater number of linked and linear pixels, typical examples of which are shown in FIGS. 24A and 24B. Alternatively, the half-tone picture image commonly involves a number of "dot patterns" that are generated periodically therein; therefore, if the local pattern is a part of the halftone image, it will observed such a periodic pattern typically shown in FIGS. 25A and 25B in the 4.times.4 local pixel matrix. As a still another case, if the local pattern does not show any one of the two image patterns, and if it has a random dot pattern (typical examples of it are shown in FIGS. 26A and 26B), judgment will be made that it must be a part of the continuous-tone image. These properties can be checked by statistically examining the actual patterns.
The image type discrimination table data to be stored in the ROM 808 may be prepared in the following manner.
First, actual character image patterns of various sizes are input as a training pattern through an optical reader such as the image sensor 11, and binary-frequency component patterns are acquired by the mentioned circuit. Hexadecimal numbers are given to the binary patterns for every 4 bits from the top as in FIG. 24; for example, "8888" for FIG. 24A, and "007F" for FIG. 24B. After various types of character images are subjected to the pattern numbering process, statistics are collected as to how many sample images are true character images. In other words, a histogram is prepared for such analysis; the statistics data should be normalized with the area of an image used for the data collection.
Subsequently, statistics on binary patterns of high frequency components is likewise collected for the halftone image. At this time, statistics data should be obtained using halftone images with different numbers of lines. FIGS. 7A and 7B, which exemplify 3.times.3 patterns as in the previous embodiment, show the patterns arranged in the order of the pattern numbers (histograms), for example. The histograms of the individual binary patterns show that easily generatable patterns differ between a character image and a halftone image.
The image type discrimination table of the ROM 808 is designed essentially based on the aforementioned property.
Our experiment has revealed the fact that halftone images such as dot-photographs are different from character/line-based images in the frequencies of occurrence of binary patterns of a high frequency component (Laplacian). For each binary pattern, therefore, the ratio of occurrence for the character/line-based images to the occurrence for the halftone images is computed. Based on the ratio of occurrence thus obtained, discrimination result is finally assigned to each of the image patterns, by using "certainty" or "similarity" levels 0 to 7. In this embodiment, the level "7" is assigned to what most looks a character, whereas the level "0" to what most appears to be a halftone image. A pattern which is generated around a character or in a region that contains nothing is called "base patterns," to which "8" or "9" is assigned. A pattern with a single dot only is one example of the base pattern.
With such a scheme, the partial patterns shown in FIGS. 24A and 24B which are discriminated as a character/line-based image (judged to be "6"). The patterns shown in FIGS. 25A and 25B are determined to be a halftone image (determined to be ("0"). The patterns shown in FIGS. 26A and 26B are judged to be "4". This means that these patterns cannot definitely be judged as any one of the character/line-based images and the halftone images, and that these patterns must be parts of a continuous-tone image such as a photograph.
The preferred embodiment of the discrimination table code data which when combined with the ROM 808 illustrated in FIG. 23 will implement the teaching of the invention is given in the Appendix. The ROM code list representing the actual storage contents of the ROM 808 has been prepared by the present inventor for every possible image pattern.
Referring to the Appendix, the left column includes 1,024 different numbers of patterns each of which is represented by four code numbers of hexadecimal addresses, such ask "0000", "0040", etc. Each hexadecimal pattern number is provided with 64 pieces of codes indicating image type discrimination results respectively; for example, "9", "8", "8", . . ., "7" for "0000". In other words, the results of decision for 64 different image patterns per row are illustrated in the Appendix; as there are 1,024 rows in total, there are a total of 65,536 (2 to the power of 16) judgment results shown.
For example, the first code "9", which is found in the first position in the first row just neighboring to the hexadecimal address "0000", represents an image type decision result of a 4.times.4 pixel matrix input pattern with hexadecimal address "0000, 0000" as shown in FIG. 27. This input pattern does not have any high-density pixels; the output code "9" indicates that such input pattern is a background image region. In other words, when a 4.times.4 binary input pattern of a high frequency component (Laplacian) is input as an address "0000, 0000" to the ROM 808 with the above table contents, it automatically outputs the code "9" as an image type discrimination result.
Similarly, a code "6" appears at the fifteenth position in the first row of this ROM table shown in Appendix. This code is prepared to represent an image type discrimination result for an input pattern addressed with "0000, 000F" as illustrated in FIG. 27. As explained above, the code "6" indicates that the input pattern can be judged at higher level as a part of a character/line-based image.
After the code table presented in Appendix is installed and stored in the ROM 808 as an "image type discrimination code table", the ROM 808 can function to automatically output a suitable code data which corresponds to an image pattern of 4.times.4 pixel matrix input thereto with specific hexadecimal addresses, and which is an image type discrimination result for the input pattern.
Successful image type discrimination is highly expectable by using the aforementioned image type discrimination process using the ROM 808 with respect to every high-frequency (Laplacian) pattern of 4.times.4 pixel matrix. This very use of the high-frequency component however causes the image type discrimination to be easily affected by possible noise. To minimize such a noise problem, and to improve the maximum accuracy of image type discrimination, the discrimination results output of the ROM 808 is subjected to several kinds of averaging effected by the circuits 810, 820, 822, 824, and 826.
Practically, a character/line-based image region and a halftone image region are not often jumbled together, and are separated from each other. Simple averaging however allows the periphery of a character/line-based or halftone image to be averaged together with the result of the judgment of a region therearound. Even if the edge of a character is correctly discriminated to be character (determined to be "7") through the first judgment by the ROM 808, averaging may lower the judged value as low as "4." Around a character in an input image such as a paper document is often a background pattern or base pattern. For the base pattern, enhancement of the high-frequency component for improvement of the image quality or filtering based on a low-pass filter characteristic to remove noise does not deteriorate the image quality With such a base pattern existing around a character, a line drawing or the like, therefore, the judgment by the ROM 808 for the edge portion of a character can be maintained by considering the base pattern around the region that has been discriminated to be a character, then averaging the result, not just by simple averaging of the judged result of the ROM 808. In other words, it is possible to perform high-frequency enhancement for the edge of or a character and output the processed results. This enhancement can be successfully effected by a specific technique that will be described below.
The discrimination result code output of the ROM 808 may be a 4-bit code having the most significant bit H3 indicating a base pattern and lower three bits H2, H1 and H0 showing the similarity level of a character, as presented in Table 5.
The code output of the ROM 808 is sequentially delayed by the series array of 1-pixel delay memories 812a to 812g included in the circuit 810. The 1-pixel, 2-pixel and 3-pixel delayed outputs of memories 812a, 812b, 812c are supplied to the switching circuit 814 serving as a first selector. The 5-pixel, 6-pixel and 7-pixel delayed outputs of memories 812e, 812f, 812g are supplied to the other switching circuit 816 serving as a second selector. The selectors 814 and 816 effect switching operation in a synchronized manner so that each selector sequentially selects one of the three inputs.
More specifically, the first selector 814 generates the 1-pixel delayed code, while the second selector 816 generates 7-pixel delayed code, whereby a couple of patterns distant by 6 pixels from each other are compared with each other. The selector 814 then generates the 2-pixel delayed code, while the selector 816 generates 6-pixel delayed code, whereby patterns distant by 4 pixels from each other are compared with each other. Thereafter, the selector 814 generates the 3-pixel delayed code, and the selector 816 generates 5-pixel delayed code, whereby patterns distant by 4 pixels from each other are compared with each other.
Assuming that each of the output codes selected by the first selector 814 is represented by "X0, X1, X2, X3", and that each output code selected by the second selector 816 is represented by "Y0, Y1, Y2, Y3". The decoder 818 generates a code indicating high-precision image type discrimination result for each pair of the code outputs of the first and second selectors 814 and 816, by utilizing a logical equation defined by an internal conversion table prestored in the decoder 818. The conversion table is illustrated in Table 6 presented below:
As understood from viewing Table 6, when the selected codes Xi and Yi (i=0, 1, 2, 3) are both base pattern (case 1 in Table 6), the 4-bit output code Zi (i=0, 1, 2, 3) of the decoder 818 is obviously a base pattern. When the second input code Yi is not a base pattern (case 2), the output of the decoder 818 is the same as the code of the second input code Yi. When the first code Xi is not a base pattern (case 3), the code of it is output directly. When neither the first code Xi nor the second code Yi is a base pattern, and if the second highest bit is "0" in other words, when both parts Xi and Yi look like a halftone image (case 4), that of their codes which has a higher similarity to a halftone image is output. As in the cases 5 and 6, when neither the first part Xi nor the second part Yi is a base pattern and the second highest bits do not coincide with each other, i.e., when one of them is a character-like pattern while the other is a pattern similar to a halftone image, nothing specific can be said about them so that they are replaced with a base pattern. Lastly, as in the case 7, when neither the first part nor the second part is a base pattern and the second highest bit is not "0," i.e., when both parts look like a character, that of their codes which has a higher similarity to a character is output by the decoder 818.
Through the above transformation, of codes to be discriminated, base image codes around a character or halftone image on an input image is replaced with a character code or a halftone image code, respectively. Accordingly, the subsequent averaging can reduce an error occurring in the judgment of the edge portion of a character. Then, the thus replaced code is subjected to averaging of 2 rows.times.8 pixels (16 pixels in total). More specifically, the replaced code is sent to the adder 822 where it is added to a code delayed by a line delay 824, the added code is then delayed by eight pixels by the series array of 1-pixel delay memories 820, and the delayed results are added and averaged by the adder 826. In this manner, the accuracy of the judgment can be improved.
A description will now be given of a still another embodiment which uses such an image recognition circuit to improve the image quality.
Referring to FIG. 28 (those elements common to the original FIG. 1 are given the same reference numerals while those different elements are denoted by different numerals), an original or an input document 2 is illuminated by a light source 1 and the image of the document 2 is formed in a CCD sensor 4 by means of an image pickup 3. The CCD sensor 4 converts the image into an electrical signal which is in turn amplified by an amplifier 5. The amplified signal is converted into a digital signal by an A/D converter 6. The digital signal undergoes unevenness compensation in a compensator 7. The operation up to this stage is essentially the same as done in the previous embodiment. Then, the optical signal is converted into a signal representing the amount of ink by a color converting circuit 12. This conversion system may also be the same as the one employed in the previous embodiment.
The quality of the image is then improved by an image improving circuit 900. Unlike the previous embodiments, this embodiment discriminates the type (property) of an input image signal, controls the high-frequency enhancement, low-pass filtering and gamma characteristic according to the discrimination signal, and sends the resultant image signal to a color printer 16, which can record multiple gradations. Through the above processing, an image such as a character will have its edge sharpened by the high-frequency enhancement while a halftone image or the like becomes a smooth image without moire noise by the low-pass filtering.
The image improving circuit 900 will be described in more detail with reference to FIG. 29. The output of the color converting circuit 12 is converted into individual ink-quantity signals, yellow (Y), magenta (M) and cyan (C). As the color printer 16 can print only one ink color at a time, the outputs of the color converting circuit 12 ar generated one color component or ink-quality signal Y, M or C at a time to match the performance of the color printer 16. The ink-quality signal is delayed by a delay circuit 902 to be a signal x and is also processed in a low-ass filter 904 to be a signal x'. The difference between the signals x and x' is then computed by an arithmetic operation unit 906 to provide only a high frequency component (x--x'), which is in turn sent to an image type discriminator 908 that discriminates the type (property) of an image as described earlier.
The high frequency component (x--x') and the delayed image signal (x) are respectively supplied to delay circuits 910 and 912. In accordance with the discrimination signal, the high frequency component is multiplied by a constant K by a multiplier 914; this may be done using a ROM table. K is variable according to the discrimination signal. For instance, when the image shows the most character-like property (when it is discriminated to be "7"), the proper K would be about 8, and when it is determined to be a halftone image (judged to be "0"), it is proper to set K=-1. This signal K(x--x') is added to the signal x in an adder 916. The resultant signal is x+K(x--x'). When K=0, this signal becomes the original signal x, and with K=-1, it becomes the output signal of the low-pass filter. When K is equal to or greater than 1, the signal becomes x+K(x--x'), which is the high frequency component (x--x') added to the original image to thereby enhance the high frequency component. In this manner, with the values of K made variable between 0 to 8 according to the signal of the discriminator 908, a high frequency component can be enhanced using the low-pass filtering that cuts off a high frequency component.
Then, the gamma characteristic is changed by the signal from the discriminator 908 to convert a character/line-based image into a clear and sharp image signal. That is, when the image is judged to show the most character-like property (when it is judged to be "7"), the gamma characteristic is made sharp by a gamma compensation circuit 918 to provide a sharp image signal as per the previous embodiment. The gamma circuit 918 may be designed to permit recording of a plurality of gamma characteristics using a ROM and select the proper characteristic according to the discrimination signal. ##SPC1##
Claims
What is claimed is:
1. An image information signal processing apparatus used in an image display system including image information input means for obtaining an image information input means for obtaining an image information signal including an input image by optically scanning the input image, and image information output means responsive to the image information signal to reproduce an output image corresponding to the input image, said apparatus for processing the image information signal so as to enable said image information output means to pleasantly reproduce the output image, comprising: image-type discriminating means for discriminating the type of the input image on the basis of a distribution, within a pixel array of a specified size, of image characteristic signals derived from image information signals each representing one pixel of the input image, wherein the image characteristic signals are each a binary Laplacian signal of an image information signal corresponding to one pixel of the input image; and means responsive to said image-type discriminating means to select one of processing for binary encoding the image information signal in accordance with the type of input image, the processed image information signal being applied to said image information output means.
2. The image information signal processing apparatus according to claim 1, wherein the Laplacian signal corresponds to the difference between the image information signal corresponding to one pixel of the input image within the pixel array and an average of the image information signals corresponding to the pixels of the pixel array.
3. The image information signal processing apparatus according to claim 1, wherein the input image includes a first type of character/line-based image, a second type of halftone image, and a third type of continuous-tone image, and wherein said binary encoding processing selecting means is responsive to said image-type discriminating means to select, with respect to the first type of character/line-based image, a processing for simple binary encoding of the image information signal which is carried out on the basis of a fixed threshold value and, with respect to the other types of images, a processing for dithering the image information which is carried out on the basis of a variable threshold value.
4. The image information signal processing apparatus according to claim 3, wherein said binary encoding processing selecting means is arranged, when the type of the input image has been discriminated by said image-type discriminating means to correspond to the first type of character/line-based image, to receive an image information signal whose high frequency components are intensified and to encode this signal by the simple binary encoding processing; when the type of the input image has been discriminated by said image-type discriminating means to correspond to the second type of halftone image, to receive an averaged image information signal and to encode this signal by the dithering processing; and, when the type of the input image has been discriminated by said image-type discriminating means to correspond to the third type of continuous-tone image, to receive a non-processed image information signal and to encode this signal by the dithering processing.
5. An image information signal processing apparatus used in an image reproducing system including an image information input means for obtaining an image information signal corresponding to each pixel of an input image, which belongs to one of specified image types, by optically scanning the input image, and image information output means responsive to the image information signal to reproduce an output image corresponding to the input image, said apparatus for processing the image information signal so as to enable said image information output means to pleasantly reproduce the output image, comprising: Laplacian calculating means connected to receive the image information signal for calculating a Laplacian thereof for each pixel of the input image to produce a Laplacian signal; image-type discriminating means connected to receive the Laplacian signal from said Laplacian calculating means for comparing a distribution, in a specified pixel array, of binary Laplacian signals with a previously stored distribution of binary Laplacian signals for each type of the specified image types, to thereby discriminate the type of the input image; image information signal compensating means connected to receive the image information signal for producing a first image information signal whose high frequency components are intensified and a second image information signal representing an average of image signals corresponding to a specified pixel array; image information signal selecting means responsive to said image-type discriminating means to select, in accordance with the discriminated type of the input image, one of the first and second image information signals and a non-processed image information signal; and binary encoding selecting means responsive to said image-type discriminating means for binary encoding the selected image information signal on the basis of a fixed threshold value upon receiving the first image information signal from said image information signal selecting means, and dithering the selected image information signal upon receiving the second image information signal or the non-processed image information signal, the binary-encoded or dithered selected image information signal being applied to said image information output means.
6. The image information signal processing apparatus according to claim 5, wherein the Laplacian corresponds to the difference between an image information signal corresponding to one pixel within a pixel array and a signal representing an average of image information signals corresponding to the pixels of the pixel array.
7. The image information signal processing apparatus according to claim 5, wherein the image information signal is a brightness signal.
8. The image information signal processing apparatus according to claim 5, wherein the specified image type include a first type of character/line-based image, a second type of halftone image, and a third type of continuous-tone image; and wherein said binary encoding selecting means is responsive to said image-type discriminating means to select, with respect to the first type of character/line-based image, a simple binary encoding processing to binary encode the selected image information signal on the basis of a fixed threshold value and, with respect to the other types of images, a dithering processing to binary encode the selected image information signal on the basis of a variable threshold value.
9. The image information signal processing apparatus according to claim 5, wherein said image-type discriminating means includes first circuit means for comparing a distribution, in the specified pixel array, of binary Laplacian signals with a previously stored distribution of binary Laplacian signals for each type of the specified image types; second circuit means connected to receive the Laplacian signal to examine the magnitude of the Laplacian signal; feedback circuit means for storing results of previous decisions made by said image-type discriminating means; and decision circuit means for finally discriminating the type of input image in response to said first and second circuit means and said feedback circuit means to cause said binary encoding selecting means to select either the binary encoding processing or the dithering processing.
10. An image information signal processing apparatus used in an image reproducing system including image information input means for obtaining an image information signal corresponding to each of pixels of an input image, which belongs to one of specified image types, by optically scanning the input image; and image information output means responsive to the image information signal to reproduce an output image corresponding to the input image, said apparatus for processing the image information signal so as to enable said image information output means to pleasantly reproduce the output image, comprising: circuit means for providing a brightness information signal and first and second color difference signals which correspond to each of pixels of an input image; Laplacian calculating means connected to receive the brightness signal for calculating a Laplacian thereof for each pixel of the input image to produce a Laplacian signal; image-type discriminating means connected to receive the Laplacian signal from said Laplacian calculating means for comparing a distribution, in a specified pixel array, of binary Laplacian signals with a previously stored distribution of binary Laplacian signals corresponding to each type of the specified image types, to thereby discriminate the type of input image; brightness information signal compensating means connected to receive the brightness information signal for producing a first brightness information signal whose high frequency components are intensified and a second brightness information signal representing an average of image signals corresponding to pixels in the specified pixel array; brightness information signal selecting means responsive to said image-type discriminating means to select, in accordance with the discriminated image type, one of the first and second brightness information signals, and a noncompensated brightness information signal; color difference signal compensating means connected to receive the first and second color difference signals to produce first and second average color difference signals each representing an average of color difference signals corresponding to the pixels in the specified pixel array; color conversion means connected to receive the brightness information signal from said brightness information signal selecting means and the first and second average color difference signals from said color difference signal compensating means for converting these signals into the first to third color information signals suitable for the reproduction of the output image; and binary encoding selecting means connected between said color conversion means and said output means and responsive to said image-type discriminating means for binary encoding the color information signals on the basis of a fixed threshold value when the first brightness information signal is selected by said brightness information signal selecting means, and to dither the color information signals when the second brightness information signal or the non-compensated brightness information signal is selected, the binary-encoded or dithered color information signals being applied to said image information output means.
Patent Citations (17)
| Patent | Date | Inventor | Cited By |
|---|---|---|---|
| US4194221 | 1980-03-01 | Stoffel | |
| US4288821 | 1981-09-01 | Lavallee et al. | |
| US4403257 | 1983-09-01 | Hsieh | |
| US4433346 | 1984-02-01 | Stoffel et al. | |
| US4547811 | 1985-10-01 | Ochi et al. | |
| US4554593 | 1985-11-01 | Fox et al. | |
| US4577235 | 1986-03-01 | Kannapell et al. | |
| US4663662 | 1987-05-01 | Sekizawa et al. | |
| US4811239 | 1989-05-01 | Tsao | |
| US4879753 | 1989-11-01 | El-Sherbini | |
| US4926251 | 1990-05-01 | Sekizawa et al. | |
| US5014124 | 1991-05-01 | Fujisawa | |
| US5016118 | 1991-05-01 | Nannichi | |
| EPX41400 | 1981-12-01 | ||
| EPX100811 | 1984-02-01 | ||
| JPX57-125580 | 1982-08-01 | ||
| JPX58-3374 | 1983-01-01 |
Non-Patent Literature (2)
- Schindler, H. R. et al., "Optical Scanning of Continuous-Tone and Line Documents with Spatial Frequency Separation for Improved Data Handling and Compaction", IBM Technical Disclosure Bulletin, vol. 20, No. 12, May 1978, pp. 5416-5421.
- Dashiell, Stephen R., "Digital Halftone Screener and Rescreener", Xerox Disclosure Journal, vol. 5, No. 1, Jan./Feb. 1980, p. 117.