Digital Photofinishing System Including Scene Balance, Contrast Normalization, and Image Sharpening Digital Image Processing

This invention relates in general to digital photofinishing systems and more particularly to a digital photofinishing system including scene balance, contrast normalization, and sharpening digital image processing.

Several problem areas need to be addressed when making color negative paper prints from color negative film images. The object of the process is to make a pleasing print from the film image. The first problem is to find the level of exposure necessary in a projection printer system to produce that pleasing print. In the simplest implementation of a process, one projects the color negative film image onto a photosensitive paper image receiver, processes the paper, and then repeats the process until a pleasing print has been obtained. In automated photo-finishing operations, a film scanner reads densities from the negative and passes the information to a computer algorithm that computes the appropriate exposure values so that a pleasing print can be obtained. At this point, the conventional photographic printing process ends. That is, if there are any remaining problems in a photographic image, no other simple processes are available to reduce the severity of the problem. Problems such as inappropriate gamma for a particular scene, poor quality camera or printer lenses, low or high activity of film or paper processes, poor scene balance, and poor sharpness in the final image remain that we would like to correct or modify.

Methods and systems have been described for more than 10 years that are devoted to producing pictorial images on various media and devices from scenes captured on photographic film, via scanning to produce a digital image, image processing, and output rendering. Examples include the following.

Journal of Imaging Technology, Vol. 14, Number 3, June 1988, Firth et. al. describe systems that capture scenes on film, scan film to produce a digital image, digitally process the image, and output via a laser AgX printer.

U.S. Pat. No. 4,500,919, Schreiber discloses an image reproduction system that scans an image captured on film, displays the image on a video monitor, enables image processing, and finally output to an inked hardcopy.

U.S. Pat. No. 4,979,032 (Dec. 18, 1990, filing date: Dec. 27, 1988), Alessi et al. describe an apparatus, including a film scanner, an video monitor, image processing, and output, to produce various output visually matched to the image displayed on the monitor.

U.S. Pat. No. 5,267,030, issued Nov. 30, 1993, inventors Giorgianni et. al. describe method and means to transform images captured on film, via digitization on a film scanner, to a color metric or other space, with output onto a variety of media and devices. This document describes the improvements offered by digital image processing, including aesthetically pleasing modifications to the tone and color reproduction as well as sharpening.

U.S. Pat. No. 5,300,381, issued Apr. 5, 1994, inventors Buhr et. al. describe a pictorial imaging system that consists of capture on photographic film, film scanning to produce a digital image, image processing, and digital output.

U.S. Pat. No. 5,579,132, issued Nov. 26, 1996, inventor Takahashi describes an image processing system devoted to storing or producing images that have "substantially the same color" or additional "aesthetic color correction" versus the original scene, based on a variety of image processing transformations of the digitized image.

U.S. Pat. No. 5,608,542, issued Mar. 4, 1997, inventors Krahe et. al. describe a system that produces index prints based on scanning a film frame, image processing, and rendering.

U.S. Pat. No. 4,945,406, issued Jul. 31, 1990, inventor Cok, describes a system for achieving automatic color balancing of color images by transferring color pixel values from log exposure RGB color values into printing density values and generating color correction offset values utilizing a printing density based color correction method.

The KODAK 35 mm/24 mm color negative film format Index Printer, sold by Kodak, produces an index print (a matrix of small imagettes) reproduced from individual film image frames. The index print is produced by the photofinisher when the original print order is processed and is supplied to the customer as a convenient means of identifying image frames on the film (see: U.S. Pat. No. 5,608,542, above). The Kodak Index Printer uses image processing on the miniature images including:

digital image in film RGB printing density

applying a scene balance algorithm to balance the digital film printing density image

mapping the color negative digital image onto a color paper (EDGE-type) characteristic curve

digital sharpening

rendering using a CRT printer onto photographic paper

In the Index Printer, the above image processing: (1) is not applied to full frame images in a digital color printer; and (2) is not applied to high resolution images in a digital color printer, only low resolution images.

U.S. Pat. No. 5,012,333, issued Apr. 30, 1991, inventors Lee et al., discloses a dynamic range adjustment system for printing digital images based on visual photoreceptor adaption and human visual contrast sensitivity. The system adjusts the contrast of the low frequency component only of the image, preserving the high frequency component in its contrast.

It is also possible to improve the tone reproduction of a captured scene using another approach. The captured image contrast is first estimated by (1) forming a Laplacian histogram, (2) determining, from the Laplacian histogram, the first and second thresholds which eliminate substantially uniform areas or a substantially textured portion of the digital image, (3) selecting pixels which are based on the first and second thresholds from the digital image, (4) forming a histogram from the sampled pixels, (5) computing a standard deviation of the sampled histogram, and (6) estimating the contrast of the digital image by comparing the computed standard deviation with a predetermined contrast for determining contrast of the input image in relationship with the predetermined contrast. The captured scene contrast estimate can be adjusted to be that of an aim scene contrast, forming a correction function, and the tone scale of the captured scene can be remapped using this correction function.

U.S. Pat. No. 4,731,671 teaches a method for contrast adjustment in digital image processing. This method creates a plurality of predetermined Laplacian response intervals and then computes the Laplacian for each pixel in an input image. The code value for each pixel in the image is then placed in its corresponding Laplacian response interval for accumulating several code value histograms, one for each Laplacian response interval. It then selects the histogram whose shape is closest to a normal distinct standard deviation of the selected histogram is taken as an estimate of the scene contrast. The standard deviation is used as an estimate of the scene contrast because of the correlation between the standard deviation and the scene contrast. That is, a high standard deviation corresponds to a high scene contrast and a low standard deviation corresponds to a low scene contrast.

The estimated contrast is then compared with a distribution of scene contrasts (e.g., standard deviations from a plurality of scenes) pre-computed from a plurality of random sample of images. If the estimated contrast is higher than the population average, then the image is considered to have a higher than normal contrast and the system reproduction contrast is then adjusted lower so that the printed image will have an image contrast that is closer to the average contrast. If the estimated image contrast is lower than the average, then the system reproduction contrast is raised accordingly.

The method of U.S. Pat. No. 4,654,722 seems to perform rather well for most images; however, there are a few situations when there are shortcomings. First, there are two parameters that have to be predetermined: the lower threshold and the width of the Laplacian interval. The patent does not provide an automatic method for setting these two parameters. As a consequence, for some images, the lower threshold is not high enough to exclude noise and textures, therefore, causing the standard deviation of the selected histogram to be unduly biased by large uniform areas (when the noise is higher than the lower threshold) or by busy texture areas (such as grass or trees). Secondly, the selected histogram often exhibits bimodality for overcast scenes with sky in them. Although the scene contrast is low, the standard deviation of the selected histogram is large because of the bimodality caused by the dark grass pixels and the bright sky pixels.

A theoretical construct is described by W. A. Richards, "Lightness Scale from Image Intensity Distribution," Applied Optics, 21, 14, pp. 2569-2582, 1982. The idea is that a "randomly" sampled log-exposure histogram of an image should have a shape similar to a normal distribution.

In addition, the above-described histogram transformation method does not have a sound theoretical foundation, and frequently, it produces unacceptable tone reproduction for consumer images. Still further, in the method of W. A. Richards, a drawback arises in defining what constitutes a truly random sampling. Furthermore, another drawback of the histogram modification approach in the prior art is that the resulting tone transformation curve often has too high or too low local contrast in some portions of the curve. Therefore, the processed image does not look pleasing.

All of these articles or patents describe, in one form or another, processes for obtaining more pleasing prints from a film image capture than the conventional optical process. There is thus a need for a solution to these problems, which can be incorporated into a digital photofinishing system.

According to the present invention, there is provided a solution to the problems of the prior art.

According to a feature of the present invention, there is provided a method of digital photofinishing comprising the steps of: producing a digital color image in optical printing densities of a color image captured on photographic media; processing the digital image with a scene balance algorithm to produce a balanced digital color image; using a contrast normalization algorithm, remapping the balanced digital color image to a preferred tone reproduction position; mapping the contrast normalized digital color image through a hard copy media characteristic curve to produce a digital image mapped to print densities of the hard copy media; sharpening the mapped digital color image with a sharpening algorithm optimized to avoid unacceptable artifacts; and digitally printing the sharpened digital color image onto hard copy media.

According to another feature of the present invention, there is provided a method of digital photofinishing comprising the steps of: producing a digital color negative image in optical printing densities of a color image captured on a color negative; processing the digital color negative image with a scene balance algorithm to produce a balanced digital color negative image; using a contrast normalization algorithm; remapping the balanced image to a preferred tone reproduction position; mapping the remapped digital color negative image through a hard copy media characteristic curve to produce a digital positive color image; sharpening the mapped balanced digital color positive image with a sharpening algorithm optimized to avoid unacceptable artifacts; and digitally printing the sharpened digital color positive image onto hard copy media.

The invention has the following advantages.

1. A digital photofinishing system is provided that produces high quality digital photographic reflection prints from color negative film images at high print rates.

2. Sharpness that is difficult to correct in conventional optical photofinishing systems is corrected by digital image processing in a digital photofinisher.

3. Prints produced by the digital photofinishing system of the invention were preferred over prints produced by optical photofinishing systems.

4. Prints produced by the digital photofinishing system that have a preferred image contrast position relative to a traditional optical print.

5. The contrast normalization algorithm improves quality of low contrast scenes by increasing image contrast and of high contrast scenes by decreasing image contrast.

6. Improves quality of prints produced from low activity film and/or paper chemical processes by increasing the print contrast.

7. Improves quality of prints produced from high activity film and/or paper chemical processes by decreasing the print contrast.

FIG. 1 is a block diagram of a digital photofinishing system incorporating the present invention.

FIG. 2 is a flow diagram of a preferred method of the present invention.

FIG. 3 is a graphical view useful in explaining aspects of the present invention.

Referring to FIG. 1, there is shown a block diagram of a digital photofinishing system 10 incorporating the present invention. As shown, system 10 includes a digital scanner 12 for scanning color photographic media 14 to produce a digital color image of a color image frame 16 on color media 14. Color media 14 is a negative film. Image data manager 18 processes the digital image to optimize the image for printing the processed image by printer 20 on color hard copy media 22. Color hard copy media can be any high quality reflection or transmissive media, such as silver halide color paper, and media used in ink jet, thermal, electrophotographic printing processes. The operation of scanner 12 and printer 20 are described in greater detail below.

Image data manager (IDM) 18 processes the digital color image from scanner 12. IDM 18 is preferably a digital computer having user input 24 (keyboard, mouse) and computer readable storage media input 26. Computer readable storage media may comprise, for example, magnetic storage media, such as a magnetic floppy disc or magnetic tape; optical storage media, such as optical disc, optical tape or machine readable bar code; solid state electronic storage devices, such as read only memory (ROM) or random access memory (RAM); or any other physical device or medium employed to store a computer program. The digital image processing techniques described below can be readable storage media. Alternatively, some or all of the techniques may be incorporated into programmable gate arrays or other hard electronic devices.

Scanner

The scanner 12 for the production of digitally processed prints scans a 35 mm full frame image at a minimum resolution of m.times.n pixels. It would be preferred that the scanner produce a higher image resolution of, e.g., 2m.times.2n pixels so that "panoramic" images can be printed at a minimum 1000 pixels on the picture height axis without interpolation. The higher resolution is also preferred for the preparation of 5R prints. All magnifications higher than 5R will require higher resolution scans.

The scanner delivers the digitized data to the image processing algorithm as "printing density." A scanner that measures printing density has red, green, and blue effective spectral response that matches that of photographic paper in an optical printer. Photographic films are designed based on the expectation that the captured information in the film will be read by this type of red, green and blue spectral response characteristic.

Even though the scanner is not considered part of the digital image processing path, some manipulations of the data may be required to deliver "printing density" to the processing algorithm. Two key steps are converting scanner densities to calibrated scanner densities, and converting the calibrated scanner densities to printing densities. Matrix operations may be required to perform these conversions.

The first step in the process is to convert the raw scanner numbers to calibrated scanner densities. If a matrix operation is used, an appropriate matrix correction model is shown below. ##EQU1## N.sub.rsx is raw scanner number and D.sub.sx is corrected scanner density.

The next process in the scanner converts calibrated scanner densities to calibrated printing densities. The process is implemented again with a matrix model that in this case can be up to a 3.times.10 matrix multiplication. An example is shown below. ##EQU2## The output of the matrix multiplication is calibrated printing density. D.sub.sx is calibrated scanner density, and D.sub.px is calibrated printing density. Either or both of these steps can be implemented with a 3D look-up-table.

According to a preferred method of the present invention as shown in FIG. 2, an image on photographic media (block 30) is scanned (by scanner 12) to produce a digital image (block 32). After conversion to printing density, the digital image is sequentially processed by IDM 18 in an image processor section 33 with a scene balance algorithm (SBA) (box 34) (See: Modern Exposure Determination for Customizing Photofinishing Printer Response," J. Appl. Photog. Eng., Vol. 5, No. 2, Spring 1979, pp. 93-104), and a contrast normalization algorithm (CNA) (box 35). In a print section 37, the SBA and CNA digital image is processed with print media characteristic curve mapping (box 36), image sharpening (box 38) and then digitally printed on hard copy media (box 40) by digital printer 20.

Hard copy media includes media optimized for the print technology used, i.e., laser or CRT photographic printers, ink jet printer, thermal printer, electrophotographic printer, etc.

The scene balance algorithm and contrast normalization algorithm are used in an analytical subsystem 41 on a full or subsampled digital image (boxes 42, 44, and 46) to produce density shift parameters and a tone scale look up table that are applied to the full resolution digital image in boxes 34 and 36, respectively.

Analytical Subsystem (Box 41, FIG. 2)

The image process begins with the digitized film image. The image, with a minimum resolution of m.times.n pixels, may be sampled down to a m'.times.n' pixel image that is analyzed by a scene balance algorithm (SBA), and a contrast normalization algorithm. This information is passed to the processor section of the algorithm so that the full resolution image, or any other subsampled image, can be processed by an equivalent process.

The analysis phase determines a set of shift values, to be added to the digitized pixel values, that are necessary to assure that the final print densities will be appropriate. The scene balance algorithm determines a set of densities that represent a form of "average densities" that are compared to a known set that produces a pleasing print. These numbers force the digitized image to print to the proper print density and color balance. The average densities are subtracted from the assumed gray density point of the paper look-up table to give shift values that will modify the image to produce a pleasing print for density and color balance.

Finally, the sub-sampled image or full resolution image is analyzed using the contrast normalization algorithm. The algorithm produces a look-up table that is used to modify the printing density values of the image in a way that produces final print images with a more pleasing tone reproduction. The look-up table does the tone correction in a luminance-chromanance space. The tone scale modification is shown as a two step process; first is the rotation to luminance-chromanance space followed by the actual algorithm that produces the look-up table.

The analytical section produces two items that are sent to the processor section of the image processing algorithm; the computed shift values from the scene balance algorithm and the image dependent tone scale look-up table produced by the contrast normalization algorithm.

Scene Balance Mapping

The process of operating a two stage film/media printing system has as its most imposing task that of finding a density value on the film exposure and mapping the densities on that negative to the media so that the best overall density and color balance are obtained. The process is best illustrated by tying together a series of reference points showing the connections from a scene object to the photographic reproduction of the object.

FIG. 3, below, illustrates the concept of density tie points in our analysis of printing process. This process ties an initial target, usually a uniform gray card, to the photographic reproduction of that target. Test points are established for the initial exposure of the target object as an expected reference film density from a film reference exposure for the test object. The quadrant labeled "Printer" merely reflects the film reference density to the log exposure axis of the media. The test target will be assigned a density to be achieved for the reproduction of that object, or a density at the paper balance point. This point is characterized as the media reference exposure value and a media reference density value.

The objective of the scene balance process is to first determine the film density value for the film reference exposure and to determine the difference between this film density value and the media reference exposure value. The difference, or .DELTA., is added to the film reference density. This process assures that the quadrant labeled "Print Through" will yield the appropriate print density in the final reproduction of the scene.

Real photographic opportunities, or images, usually do not contain test objects that can be mapped in a formal process like that just described. Scene balance algorithms are designed to perform the task of estimating the reference film density, as if a gray card had been included in the scene. Once an estimate of the film reference density has been made, the process is duplicated just as described above. The examples show only one color record. Color films contain three records to capture the red, green, and blue information from a scene. Thus, the scene balance algorithms must perform two tasks. First estimate the overall density of the film image to that the best neutral density is obtained in the final print, or print through quadrant of the scene. Then the balance between the red, green, and blue exposure must be estimated.

Image Processor Section (Box 33, FIG. 2)

In box 34, the scene balance shift values determined in box 44 are applied to the printing density values. These values represent the shift (add to the red, green, and blue pixel values) necessary to achieve a print that has the correct color and density balance.

Next, the tone scale look-up table from the analysis process (box 46) is applied to the full digital image. The image must first be rotated to a luminance-chromanance space since the best tone scale results (image quality) are achieved when done on the luminance information. The matrix t.sub.-- space.mat is used in this process. (See: Modem Exposure Determination for Customizing Photofinishing Printer Response," J. Appl. Photog. Eng., Vol. 5, No. 2, Spring 1979, pp. 93-104.) ##EQU3##

After the tone scale look-up table has been applied to the image, the image data is rotated back to printing density space using the matrix t.sub.-- space.sub.-- inv.mat.

Printer Section (Box 37, FIG. 2)

The next step in the process is to map the balanced and contrast normalized image through a hard copy media (color photographic paper) characteristic curve. The printing density values of the balanced image are mapped to the appropriate print density values for negative photographic print paper. The finished process produces an image that is color and density balanced, and in a print density metric. The photographic paper curve is balanced.

Unsharp Masking Process

The final step prior to actually printing the image is the sharpening process. An unsharp masking algorithm is applied to every image just prior to printing. The unsharp masking equation is as follows.

In this equation, D.sub.c (x) is the "sharpened" density at position x, D(x) is the starting density at position x, the recommended value of .beta. is 2.5, and the integral values at each pixel position are computed by convolving the image with the kernel, below: ##EQU4## The red, green, and blue images are all sharpened to the same level.

The recommended value for .beta. was established by adjusting the value upwards until pictures, when printed, began to appear with unacceptable levels of digital artifacts. These artifacts appeared as ringing, or halos, on edges yielding images that appear unnatural.

This value of beta is then applied to a square wave target to measure the modulation transfer function for the final processed image. The test to establish the full system MTF response is as follows. First, a target is photographed using a color negative film. The original target has a square wave pattern of approximately 40% modulation. The spatial frequency response values for each frequency of square wave pattern is 100% through all of the system visible frequencies. The film image was scanned on the digital scanner to produce a 1024.times.1536 pixel digital image for subsequent processing. This image was processed through our SBA plus sharpening path and printed. The spatial frequency response of the final print was measured with a high resolution microdensitometer, and the data analyzed using a harmonic analysis process. Table 1 lists the red, green, and blue response measured in this test (average of four samples) and represents the maximum MTF before significant artifact production occurs.

The MTF curve measured using this process represents the maximum spatial frequency of any digital processing system for Digital Photofinishing. This as the final print image is a combination of camera lens, film, scanner, algorithm, print engine, and print media. Any combination of these elements that yields and MTF curve of this result, or any result below the values listed in Table 1, will be considered as part of this invention. The set of parameters considered for this system, including the sharpening algorithm, are the maximum level of boost before going into a condition of oversharpening. Values of beta lower than that specified in the report are considered within the scope of this invention because these values also deliver prints that meet our printing requirements for making digital prints. If one of the system components is changed, then the beta value will be changed to assure that any final image produced by this digital path will achieve the upper limit MTF curve.

Psycho-physical experiments were conducted to compare these digital path prints to a convention optical printing path. The optically printed images were prepared using a CLAS 35 optical printer running in the full order printing mode, thus emulating the SBA operating in our digital path. The scene balance algorithm parameters used in the digital and optical paths were the same so that similar prints could be prepared. Pairs of prints, digital and optical, were shown to a panel of three judges who were asked to choose the best print from the pair. In 75% of the pairs, the digital print was selected, citing sharpness as the reason. The remaining 25% were the optical prints selected because of grain build-up in the digital prints.

Printer

At this point in the process, the printing densities should be fully balanced and corrected. This image information can be printed through a simple printing density to print density look-up table.

In the simplest case (printing density to print density), the printing density values are mapped to the appropriate print density values for a negative photographic paper. The finished process produces an image that is color and density balanced, and in a print density metric. We find the most preferred mapping from the balanced digital color negative image onto the characteristic curve of the aim AgX paper results from mapping reference RGB printing densities representing an achromatic middle gray onto achromatic RGB paper densities. These RGB paper densities are a function of the print material image dye spectra.

The final step prior to actually printing the image is the sharpening process which has been described previously.

At this stage in the process, the encoded data is sent to a printing device that renders, or prints, the information. A calibration process must be operating on this device such that the code values presented to the printer will yield the expected print densities.

Printer Calibration

Printer calibration will be done as part of the image processing system maintenance such that test patch density differences between measured and expected densities of less than 0.01 are obtained. The simplest embodiment only requires a neutral scale calibration. In more complex applications, a color calibration may be necessary.

A series of uniform patches (at least 18) spanning the full range of printer code values are printed through an initial calibration LUT. This initial LUT must cover all those D/A count values that produce density on the print. The patch densities on the print are measured. With the initial LUT, the list of code values, respective densities of those patches, and the aim curve, a new calibration LUT can be calculated which should modify printing behavior according to the calibration aim.

It will be appreciated that image sharpening can be effected at other points in the image processing chain. It will also be appreciated that the goal of the image processing described herein is to produce images via digital printing that are similar to, or better than, images produced by optically printing negative film images. This can be done in a straight forward manner as described herein when using digital hard copy media having similar image dye spectra to the image dye spectra of the hard copy media (silver halide color paper) used in optical printing.

In general, scene contrast modification algorithms contain adjustable parameters that provide a mechanism to improve the algorithm results for specific scenes, for specific capture media, and/or for specific customers. These adjustable parameters may be fixed for a specific software version embodiment of the algorithm. Or they may be obtained by the software embodiment of the algorithm from different sources. They may be encoded on the capture media optically or magnetically. They may be encoded on optical, magnetic, or other media that accompany the capture media. They may be selected or supplied, directly or indirectly, by the customer or processing lab. They may be accessed from a software embodiment database, based on the film type, lab preference, or customer preference. They may also be accessed via the intranet from an available repository, based on the film type, lab preference, or customer preference.

The invention has been described in detail with particular reference to certain preferred embodiments thereof, but it will be understood that variations and modifications can be effected within the spirit and scope of the invention.

10 digital photofinishing system

12 digital scanner

14 color film

16 color image frame

18 image data manager

20 printer

22 color print paper

24 input

26 computer readable storage media input

30 image on photographic media

32 produce digital image

34 process with scene balance algorithm

36 print media characteristic curve mapping

38 image sharpening

40 digitally print image

42 create subsampled digital image

44 SBA density shift parameters

46 full resolution digital image