Apparatus and Method for Treating Encoded Image Data Including Correction for Encoding/Decoding Deterioration

1. Field of the Invention

The present invention relates to a method and apparatus for decoding encoded image data, which has been encoded by, e.g., block encoding.

2. Related Background Art

When a document or an image is transmitted or stored, a redundancy reduction code is used to shorten transmission time or to effectively utilize a storage medium. Redundancy reduction encoding is roughly classified into two encoding methods. One method is called a reversible encoding method, and the other method is called an irreversible encoding method.

In the reversible encoding method, data is reversibly encoded. Therefore, loss of information, i.e., deterioration of image quality does not occur. Typical reversible encoding methods are the MH (modified Huffman) and the MR (modified READ) encoding which are used in facsimile systems. The reversible encoding method was developed to process binary data.

Irreversible encoding method include block encoding, vector quantization, orthogonal conversion encoding, and the like. The assignee of the present invention has proposed the irreversible encoding method in U.S. Pat. No. 4,797,945. However, when the irreversible-encoded data is decoded, information is normally lost, resulting in deterioration of image quality.

Deterioration of image quality in the irreversible encoding method does not occur in a region where variations in gray level or tone are small, but noticeably occurs in regions where variations in gray level are large, e.g., an edge portion of a character. As a result, an edge of a character is made unclear or blurred. In a region where variations in gray level are small, a block becomes conspicuous, and gradation visibly deteriorates.

It is an object of the present invention to provide a method and apparatus for the decoding of encoded image data, which are free from the conventional drawbacks, and can correct a degraded image the deterioration in which is due to the encoding and decoding operations, to obtain an image which does not suffer from image quality deterioration.

It is another object of the present invention to provide a decoding apparatus and method for decoding encoded image data, which correct deterioration of an image in accordance with characteristics of decoded image data.

It is still another object of the present invention to provide a decoding apparatus and method for decoding encoded image data, which selectively perform a plurality of types of deterioration correction processing for decoded image data.

The above and other objects and effects of the present invention will be apparent from the following detailed description of the preferred embodiment.

FIG. 1 is a block diagram of an embodiment in which a decoding apparatus according to the present invention is applied to a printing apparatus;

FIGS. 2(a) to 2(f) and FIGS. 3(a) to 3(f) are views for explaining correction states of an image which suffers from image quality deterioration;

FIG. 4 shows a matrix of a spatial filter for edge emphasis;

FIGS. 5(a) and 5(b) are views showing a state wherein an edge whose image quality is deteriorated is emphasized;

FIG. 6 is a block diagram of an edge emphasis circuit;

FIG. 7 shows a matrix of a smoothing spatial filter for smoothing;

FIG. 8 is a block diagram of a smoothing circuit;

FIGS. 9(a) and 9(b) show matrices of a primary differential spatial filter for extracting a low-frequency component of image data;

FIG. 9(c) shows a matrix of a secondary differential spatial filter for extracting a high-frequency component of image data;

FIG. 10(a) is a graph showing spatial frequency characteristics of the primary and secondary differential filters;

FIG. 10(b) is a graph showing spatial frequency characteristics of an edge image and a stepped image area;

FIG. 11 is a detailed block diagram of a deterioration detection circuit for detecting image quality deterioration; and

FIG. 12 is a circuit diagram showing an arrangement of a determination circuit.

An embodiment of the present invention will be described hereinafter in detail with reference to the accompanying drawings. For the sake of simplicity, encoding used in the following embodiment is block encoding, wherein image data is divided into blocks each having a predetermined size, and a block of interest is represented by an average density of image data in the block.

FIG. 1 is a block diagram showing an embodiment in which a decoding apparatus according to the present invention is applied to a printing apparatus which receives block-encoded image data, performs image processing, and prints the processed image data. With this printing apparatus, when image data which is block-encoded by an encoding circuit 1 is transmitted by a transmission circuit 2, a receiving circuit 3 of the printing apparatus receives the image data, and a decoding circuit 4 then performs predetermined decoding. FIGS. 2 and 3 show image data which is block-encoded and is then decoded.

FIG. 2(a) shows an original image, and FIG. 2(d) shows its one-dimensional density distribution along the abscissa of FIG. 2(a). The encoding circuit 1 shown in FIG. 1 block-encodes an original image using a 4.times.4 (pixel) block unit. That is, instead of transmitting 16 (=4.times.4) pixels, an average density in a block is transmitted. In such irreversible encoding, information of a pixel having a density different from the average density is lost. Therefore, even if the received image data is decoded by the decoding circuit 4, if an original image has an edge, its sharpness may be lost, as shown in FIGS. 2(b) and 2(e).

FIG. 3 shows another original image in which gray levels moderately change. In this case, a block encoding unit is also a 4.times.4 (pixel) block for the sake of simplicity. When image data which is encoded by an average density of a 4.times.4 block is decoded, its density is changed stepwise, as shown in FIGS. 3(b) and 3(e). In this case, moderate changes in gray levels from one part of the block of the original image to the next is also lost.

In view of the above, image quality deterioration is corrected by a post-processing circuit 100 as the characteristic feature of this embodiment. More specifically, the post-processing circuit 100 performs edge emphasis on an image whose edge property is lost using an edge emphasis circuit 6, and performs smoothing for an image which is deteriorated as a result of having a stepped change in density using a smoothing circuit 7, thereby correcting the image. The corrected image data is output to a known printer 13 (or a high-resolution CRT apparatus), and is reproduced with high reproducibility and high quality. The post-processing circuit 100 will be described below. The post-processing circuit 100 performs image processing such as edge emphasis and smoothing. For this purpose, the post-processing circuit 100 comprises a line memory 5 for extracting a 3 .times.3 block from decoded image data, a deterioration detection circuit 100b for detecting image quality deterioration of the extracted block, and a correction circuit 100a for performing appropriate correction in accordance with a detection result of image quality deterioration of the block of interest.

The correction circuit 100a comprises the edge emphasis circuit 6, the smoothing circuit 7, and a selector 12 for selecting one of an output 6a from the edge emphasis circuit 6, an output 5a from the line memory 5, and an output 7a from the smoothing circuit 7 in accordance with a detection result 11a from the deterioration detection circuit 100b.

The deterioration detection circuit 100b comprises a density difference detection circuit 8 for detecting a density difference in a 3.times.3 block extracted by the line memory 5, two spatial filters 9 and 10 for extracting high- and low-frequency components of a spatial frequency of the 3.times.3 block, and a determination circuit 11 for determining image quality deterioration from the density difference and the extracted spatial frequency.

FIG. 4 shows a matrix of a Laplacian spatial filter used for edge emphasis This matrix has a size of 3.times.3 corresponding to a block extracted by the line memory 5. Nine pixel elements X.sub.11 to X.sub.33 are output from the line memory 5. Of these pixel elements, X.sub.22 is here termed the pixel of interest in a block. In this case, matrix elements are X.sub.22 =5, X.sub.11 =-1, X.sub.13 =-1, X.sub.31 =-1, and X.sub.33 =-1. With this Laplacian filter, an edge portion of an image which is decoded as shown in FIG. 5(a) is emphasized and output as shown in FIG. 5(b). In the case of FIG. 2, an image as shown in FIGS. 2(c) and 2(f) is obtained.

FIG. 6 is a detailed block diagram of the edge emphasis circuit 6. In this edge emphasis circuit 6, pixel densities of a block corresponding to X.sub.22, X.sub.11, X.sub.13, X.sub.31, and X.sub.33 in the Laplacian matrix are input from line memory 5 to multipliers 100 to 104. The outputs which are multiplied with corresponding coefficients of the matrix from the multipliers are added to each other by an adder 105. More specifically, the output 6a from the adder 105 is as follows:

The output 6a is input to the selector 12.

FIG. 7 shows a matrix of a spatial filter used in the smoothing circuit 7. All the coefficients in this matrix are 1/9, and a position of a pixel of interest corresponds to X.sub.22. FIG. 8 shows an arrangement of the smoothing circuit 7. An adder 150 adds image densities X.sub.11 to X.sub.33 output from the line memory 5, and a divider 151 performs division by 9. The resultant output 7a is as follows: ##EQU1## The resultant image is input to the selector 12, and is selected in accordance with a value of a deterioration detection signal 11a. In this case, an image having a stepped change in density as shown in FIGS. 3(b) and 3(e) is smoothed as shown in FIGS. 3(c) and 3(f), and the resultant image is similar to an original image.

An operation of the deterioration detection circuit 100b for performing image quality deterioration determination will be described below. Image data X.sub.11 to X.sub.33 extracted by the line memory 5 are input to the edge emphasis circuit 6, the smoothing circuit 7, and the like, and the same image data are synchronously input to the density difference detection circuit 8, and spatial filters 9 and 10.

The density difference detection circuit 8 calculates the absolute value of a difference between an average value in the 3.times.3 block and the pixel of interest (X.sub.22). In general, if the absolute value of the difference is smaller than a given threshold value, it can be considered that the pixel of interest is present in an area in which gray levels change relatively moderately. In contrast to this, if the absolute value is larger than the threshold value, it can be considered that an image is a character image with a blurred edge or an image having a stepped change in density due to block encoding, as shown in FIGS. 3(b) and 3(e). Therefore, when the density difference detection circuit 8 determines that the pixel of interest is present in an image area in which a change in density is small, the circuit 8 outputs data "0". When the circuit 8 determines that the pixel of interest is present in a character image area with a blurred edge (FIGS. 2(b) and 2(e)) or a stepped image area (FIGS. 3(b) and 3(e)), the circuit 8 outputs data "1".

The first spatial filter 9 constitutes a spatial filter for performing edge detection for an image having a relatively low frequency (about two line pairs lp/mm). For example, the spatial filter can be realized by combining primary differential filters having, e.g., the coefficients shown in FIGS. 9(a) and 9(b) in a matrix. The second spatial filter 10 constitutes a spatial filter for detecting a high-frequency component of a spatial frequency generated in image data decoded by the decoding circuit 4. The spatial filter can be realized by a secondary differential filter as shown in FIG. 9(c). According to general filter characteristics, if the primary and secondary differential filters have the same kernel size, the primary differential filter has a peak of a transmission range characteristic at a lower frequency side than that of the secondary differential filter. FIG. 10(a) illustrates characteristics of the two types of spatial filters using one-dimensional models. The primary differential filter is selected such that a peak appears at about 2 lp/mm. The secondary differential filter is selected such that a peak appears at about 4 lp/mm.

FIG. 10(b) shows spatial frequency characteristics of a character area with a blurred edge and a stepped image area in a decoded image. In terms of frequencies, in a line image such as a character with a blurred edge, the power spectrum is continuously changed from a low-frequency side toward a high-frequency side. However, in a stepped image as shown in FIGS. 3(b) and 3(e), peaks of the power spectrum discontinuously and periodically appear at high-frequency components corresponding to a cycle of the step.

Therefore, since the transmission range is present at 2 lp/mm, the first spatial filter 9 can detect a spatial frequency component of a character edge, but cannot detect a spatial frequency component of an edge portion of a stepped image. However, in the second spatial filter 10, the transmission range is present at 4 lp/mm and the half-width of a peak representing the filter characteristics is wide. Therefore, the filter 10 can detect a frequency component of a character to have a spatial frequency of the stepped image as a center.

FIG. 11 is a detailed block diagram of the deterioration detection circuit 100 (FIG. 1) for determining an image area in an image. In the block diagram of FIG. 11, the density different detection circuit 8 comprises an average value calculation circuit 51 and a subtractor 52. The average value calculation circuit 51 receives one block of pixel data from the line memory 5, and performs addition and subtraction to calculate an average value in the block. One input terminal of the subtractor 52 receives the density of a pixel of interest (the central pixel if the 3.times.3 matrix is employed). The subtractor 52 calculates the absolute value of a difference between the average value and the density of the pixel of interest.

The absolute value of the difference is compared with a threshold value T1 by a comparator 57. If the difference between the image data of the pixel of interest and the average value is smaller, the comparator 57 determines that the pixel of interest is present in an area where image quality deterioration is low, and outputs data "0". In this case, the selector 12 selects the direct output 5a from the line memory 5. If the difference is larger, the comparator 57 determines that the pixel of interest is present in a character area with a blurred edge or a stepped image area, and outputs data "1". An output result A is input to the determination circuit 11.

One block of image data from the line memory 5 is input to the first spatial filter 9, i.e., to primary differentiators 53 and 54 shown in FIG. 11. The differentiators 53 and 54 perform the primary differential calculations shown in FIGS. 9(a) and 9(b), and output the absolute values of the calculation results, respectively. Each differentiator comprises multipliers for multiplying pixel densities with matrix coefficients corresponding to the pixels, and an adder for adding the products. Thus, the differentiator 53 detects an image edge at a relatively low frequency in the vertical direction, and the differentiator 54 detects a low-frequency edge in the horizontal direction. These outputs are added to each other by an adder 56, and the sum is compared with a predetermined threshold value T2 by a comparator 58. If the sum is larger than the threshold value T2, the comparator 58 determines that the low-frequency edge is present, and outputs data "1". If the sum is smaller, the comparator determines that no edge is found, and outputs data "0". An output result B is input to the determination circuit 11 as in the result A.

In the second spatial filter 9, a secondary differentiator 55 receives pixel data in a block, performs a secondary differential calculation shown in FIG. 9(c), and outputs the absolute value of the calculation result. The absolute value is compared with a predetermined threshold value T3 by a comparator 59. If the secondary differentiated value is larger than the threshold value T3, the comparator 59 determines that a high-frequency edge component of a stepped image is present, and outputs data "1". If the resultant value is smaller, the comparator 59 determines that no edge is present, and outputs data "0". A result C is input to the determination circuit 11.

(where X indicates either "0" or "1" is possible)

When the output A is "1" and the outputs B and C are "1", as shown in Table 1, the determination circuit 11 determines a character area with a blurred edge. If the output B is "0" and the output C is "1", the circuit 11 determines a stepped image. The determination circuit as 11 may comprise a simple circuit shown in FIG. 12.

A signal obtained by inverting the output A through an inverter 70 is used for indicating an area where image quality is not deteriorated (for selecting the direct output from the line memory 5), and the outputs from AND gates 71 and 72 are respectively used for indicating a character area with a blurred edge portion (for selecting the output from the edge emphasis circuit 6) and for indicating a stepped image (for selecting the output from the smoothing circuit 7). The selector 12 shown in FIG. 1 is switched in accordance with these indication signals. More specifically, image data output from the selector 12 is image data whose image quality deterioration is corrected. Therefore, an image reproduced by the printer 13 is a high-quality image. In particular, an image which is enlarged at a work station having an enlargement function has noticeable image quality deterioration. In this case, the above embodiment is very effective.

In the above embodiment, edge emphasis and smoothing are performed by the correction circuit 100a in accordance with image quality deterioration. However, either one of the above image corrections may be performed depending on encoding methods.

If the printing apparatus is a facsimile system, the encoding circuit 1 is arranged in a reception-side facsimile, and the receiving circuit 3 is arranged in a transmission-side facsimile. When corrected image data is stored in a data storage apparatus instead of outputting a corrected image to the printer 13, the transmission circuit 2 and the reception circuit 3 are connected through a bus line.

In the above embodiment, block encoding has been exemplified. Therefore, image quality deterioration is determined in correspondence with the block encoding. If other irreversible encoding methods are employed, a transmission frequency of the filters need only be changed.

The image correction is not limited to edge emphasis and smoothing. For example, the present invention may be applied to correction such as interpolation for interpolating a lost pixel.

According to a decoding apparatus of encoded image data of this embodiment as described above, an image whose image quality is deteriorated due to irreversible encoding can be corrected to be an image free from image quality deterioration.

The preferred embodiment of the present invention has been described. However, the present invention is not limited to the above embodiment, and various changes and modifications may be made within the spirit and scope of the appended claims.