JP4661659B2 - Photo image processing apparatus and photo image processing method - Google Patents

Description

  The present invention is capable of reproducing natural colors when performing monitor or photographic printing on color image data obtained by reading a photographic film such as a negative film or color image data photographed with a digital still camera or the like. The present invention relates to a photographic image processing apparatus and a photographic image processing method for adjusting the color balance of R (red), G (green), and B (blue) (hereinafter referred to as “RGB”).

  In a conventional photographic printer, a LATD (Large Area Transmission Density) exposure method based on Evans's theorem is known as a photographic image processing method for printing an image recorded on a negative film on a photographic paper which is a photosensitive material with good hue. Yes. This exposure method is based on Evans's theory that an average outdoor subject will be close to gray when the colors of the negatives are mixed together. In this method, exposure is performed by adjusting each RGB exposure amount so that light is reproduced in gray on a photographic paper.

  Specifically, the negative film is irradiated with light, the transmitted light is read by the image sensor to generate RGB color image data, and the average pixel value for each RGB component for all the constituent pixels is calculated and derived. The analog photo printer adjusts the dimming filter to expose the photographic paper so that each value is a predetermined value corresponding to gray, and the digital photo printer adjusts the exposure from each RGB light source. Was.

  According to the above-described conventional photographic image processing method, there is a problem that an overcorrected photographic print is output due to the color deviation of the subject (person, background). For example, in the case of a scene where a person is photographed against a lawn, the lawn area is finished in gray, while magenta, which is a complementary color of the lawn, appears strongly in the person area. Such a situation is called a color feria, and as a countermeasure therefor, a method of removing a high saturation pixel in the LATD exposure method or obtaining a conditional average value weighted by the saturation has been proposed.

JP 2000-330221 A

  However, according to the above-described method, in the case of a scene with a large color deviation, the number of pixels actually used in the calculation tends to be less stable, and even a small weighting corresponds to this. When the number of pixels to be processed is large, there are cases where the influence is not small.

  Furthermore, the threshold for removing high-saturation pixels and the weighting condition based on saturation are determined based on experience in consideration of various situations at the time of shooting such as the light source type, season, and time zone. The photographic image taken by the camera is not always perfect because of the influence of the difference in film type and development state, so there is room for further improvement.

  In view of the above-described conventional drawbacks, an object of the present invention is to provide a photographic image processing apparatus and a photographic image capable of correctly detecting a color feria degree and obtaining an appropriate color correction value for an input image having various conditions. It is in providing a processing method.

In order to achieve the above-described object, the first characteristic configuration of the photographic image processing apparatus according to the present invention is that, as described in claim 1 of the claims, the colors for all the constituent pixels in the input frame image unit. Mean pixel value calculating means for obtaining an average pixel value for each component, and pixels of the hue saturation coordinate system in which each pixel of the frame image has a saturation defined by a distance from the center and a hue defined by an angle around the center A plurality of concentric circles centered on a pixel obtained by converting a pixel represented by an average pixel value for each color component into the hue saturation coordinate system in the hue saturation coordinate system; A circular area characteristic value calculating means for defining a circular area characteristic value for each area, defining a circular area characteristic value indicating a distribution characteristic of pixels included in each circular area, and a first correction for determining a color correction amount based on the circular area characteristic value A point comprising a quantity calculation means A.

  If there is a difference in shooting conditions including shooting light sources such as sunlight or tungsten light, or a difference in film type, the entire frame image will basically be affected, and even if the same scene is shot, The formed image has a different color appearance. However, there is no difference if attention is paid to the relative colors between objects in the frame image. Therefore, when LATD correction is performed, the characteristics of the entire frame image based on the shooting conditions, film type, and the like are corrected as an achromatic color as a whole, and only the color feria phenomenon in which the averaging is excessively applied remains as a disadvantage.

  By defining a plurality of concentric circle regions centered on the average pixel value obtained by the average pixel value calculation means in the hue saturation coordinate system and evaluating the distribution characteristics of the pixels of the frame image in each region, The distribution characteristics of each circular area are almost equal in a frame image where no image occurs, but the difference in distribution characteristics between each circular area is large in a frame image where the tendency of color feria is strong, and the situation of color feria is characteristic It is reflected in. In other words, when a cumulative histogram of the number of constituent pixels is taken from the center-side circular region to the outer circular region, a frame image with a weak tendency of color feria is saturated in a relatively low saturation region, that is, the central-side circular region. On the other hand, there is a significant difference that a frame image having a strong tendency of color feria does not saturate to a relatively high saturation region, that is, an outer circular region. This tendency applies not only to the distribution characteristic of the number of pixels but also to the pixel value.

  Therefore, a first correction amount calculation means for calculating a color correction amount so as to exhibit a predetermined distribution characteristic based on a circular area characteristic value indicating a distribution characteristic of pixels included in each circular area is provided, and the average pixel value calculation is performed. By correcting the average pixel value obtained by the means, an appropriate color correction amount that avoids the occurrence of color failure can be obtained.

In the second feature configuration, as described in claim 2, in addition to the first feature configuration described above, the circular region characteristic value includes the number of pixels included in each circular region and each circular region. The area average pixel value, which is the average value for each color component of the included pixels, or the difference value between the area average pixel value and the average value of the average pixel value for each color component .

  The distribution characteristics of the number of pixels included in each circular area and the area average pixel value that is an average value for each color component of the pixels included in each circular area, or the area average pixel value and the average pixel value Both of the distribution characteristics of the difference values are prominent parameters indicating the tendency of color feria. Accordingly, the first correction amount calculation means can calculate and derive an appropriate color correction amount with less tendency of color feria based on these parameters.

  In the third feature configuration, as described in claim 3, in addition to the first or second feature configuration described above, the first correction amount calculation means is a neural network that is learned by an error back propagation learning method. Or a point constituted by multiple regression analysis means.

  As the first correction amount calculation means, a circular region of the original image is obtained for an image obtained by performing an appropriate color correction on a large number of sample images with different shooting conditions by a manual correction process to obtain an appropriate color correction amount. Arithmetic means for obtaining a color correction amount for an arbitrary image by pattern matching based on a characteristic value and an appropriate color correction amount or a statistical method can be suitably used.

  For example, the appropriate color correction amount can be obtained by inputting the circular region characteristic value of the target image to the neural network learned by the back propagation learning method using the circular region characteristic value of the sample image and the appropriate color correction amount as a teacher signal. It is possible to obtain a relational expression between the circular area characteristic value of the sample image and the appropriate color correction amount by multiple regression analysis and input the circular area characteristic value of the target image to the relational expression. A correct color correction amount can be obtained.

In the fourth feature configuration, as described in claim 4, in addition to any of the first to third feature configurations described above, the hue correction coordinate system includes the first correction amount calculation unit . The derived color correction amount is converted into the hue / saturation coordinate system as a pixel , a plurality of radial regions are defined in the circumferential direction at predetermined angular intervals around the converted pixel, and the pixels included in each radial region are defined. A radial region characteristic value calculating unit that obtains a radial region characteristic value indicating a distribution characteristic for each region and a second correction amount calculating unit that obtains a color correction amount based on the radial region characteristic value are provided. .

The photographic print obtained by correcting the target frame image with the color correction amount obtained by any one of the first to third characteristic configurations described above is effective in suppressing color feria, but is in terms of appearance. In some cases, more appropriate correction may be required. In order to improve the appearance of a photograph, it is necessary to reproduce a white portion whiter. However, in an image corrected with the above-described color correction amount, an area close to an achromatic color but slightly colored is generated. If such a region is extracted and corrected based on a simple threshold value, there may be variations in correction in similar scenes, and whether or not correction should be made so that it becomes an achromatic color depends on the hue. . For example, when the print result is corrected to yellow or red, there is almost no problem, but when the print result is corrected to cyan or magenta, there is a high possibility of visually hindering.

  Therefore, the point where the target image is represented as an achromatic color in the hue saturation coordinate system, that is, the pixel represented by the color correction amount obtained by the first correction amount calculation means is centered on the radial region based on the hue. Color with good appearance and high quality in consideration of hue by dividing the area and obtaining an appropriate color correction amount for each hue based on the radial area characteristic value indicating the distribution characteristics of the pixels contained in each divided radial area Correction can be made.

  In the fifth feature configuration, as described in claim 5, in addition to the fourth feature configuration described above, the radial region characteristic value includes the number of pixels included in each radial region and each radial region. This is in that it is composed of weighted average pixel values for each color component obtained by multiplying the included pixels by a larger weighting factor for lower saturation pixels.

  When correcting a slightly colored area to an achromatic color, when the degree of correction is adjusted for each radiation area divided by hue, the degree of correction is adjusted by weighting the pixels present in each area with saturation. can do. That is, when the weighted average pixel value for each color component is obtained so that the lower the saturation pixel, the higher the weight, and the higher the saturation pixel, the smaller the weight, the larger the area of the slightly colored region, the larger the value. It becomes like this. Therefore, by obtaining the color correction amount using the number of pixels included in each radial area and the weighted average pixel value for each color component as a radial area characteristic value, color correction with good appearance and high accuracy in consideration of hue can be performed. It becomes like this.

  In the sixth feature configuration, as described in claim 6, in addition to the fourth or fifth feature configuration described above, the second correction amount calculation means is a neural network that is learned by an error back propagation learning method. Or a point constituted by multiple regression analysis means.

  As the second correction amount calculation means, a radial region of the original image is obtained for an image obtained by performing an appropriate color correction on a large number of sample images with different shooting conditions by a manual correction process to obtain an appropriate color correction amount. Arithmetic means for obtaining a color correction amount for an arbitrary image by pattern matching based on a characteristic value and an appropriate color correction amount or a statistical method can be suitably used.

  For example, the appropriate color correction amount can be obtained by inputting the radial region characteristic value of the target image to the neural network learned by the error back propagation learning method using the radial region characteristic value of the sample image and the appropriate color correction amount as a teacher signal. It is possible to obtain a relational expression between the radiation area characteristic value of the sample image and the appropriate color correction amount by multiple regression analysis and input the circular area characteristic value of the target image to the relational expression. A correct color correction amount can be obtained.

In order to achieve the above object, the first characteristic configuration of the photographic image processing method according to the present invention is the average pixel value for each color component with respect to all the constituent pixels in the input frame image unit as described in claim 7. An average pixel value calculating step for calculating the pixel value, and a coordinate conversion step for converting each pixel of the frame image into a pixel of a hue saturation coordinate system in which saturation is defined by a distance from the center and hue is defined by an angle around the center And defining a plurality of concentric circular regions centered on a pixel obtained by converting a pixel represented by an average pixel value for each color component into the hue saturation coordinate system in the hue saturation coordinate system, A circular region characteristic value calculating step for obtaining a circular region characteristic value indicating a distribution characteristic of pixels included in the circular region for each region; and a first correction amount calculating step for determining a color correction amount based on the circular region characteristic value. To the point configured That.

In the second feature configuration, in addition to the first feature configuration described above, the color correction amount derived in the first correction amount calculation step is added to the hue saturation coordinate system, respectively. Radial area characteristics indicating the distribution characteristics of the pixels included in each radial area by converting the hue saturation coordinate system as pixels , demarcating a plurality of radial areas in the circumferential direction at predetermined angular intervals around the converted pixels A radial region characteristic value calculating step for obtaining a value for each region, and a second correction amount calculating step for obtaining a color correction amount based on the radial region characteristic value.

  As described above, according to the present invention, a photographic image processing apparatus and a photographic image processing method capable of correctly detecting the color feria degree and obtaining an appropriate color correction value for an input image having various conditions. Can now be provided.

  Embodiments of a photographic image processing apparatus incorporating an image processing apparatus according to the present invention will be described below. As shown in FIG. 2, the photographic image processing apparatus 1 includes a photographic printer 2 that performs an exposure process on the photographic paper P based on output image data and develops the exposed photographic paper, and a developed photographic film F. A media driver 32 that reads image data from an image data storage medium M such as a memory card that stores image data taken by a film scanner 31 or a digital still camera that reads an image from the camera, a general-purpose computer as a controller 33, and the like The print data is set and inputted with the print order information for the inputted photographic image as the original image, and is provided with the operation station 3 for performing various image correction processes, and the print data edited from the original image by the operation station 3 Is output to the photographic printer 2 to produce a desired photographic print. That.

  As shown in FIGS. 2 and 3, the photographic printer 2 has two systems of photographic paper magazines 21 containing roll-shaped photographic paper P, and photographic paper P drawn from the photographic paper magazine 21 with a predetermined print. A sheet cutter 22 that cuts into a size; a back print unit 23 that prints print information such as a frame number on the back of the cut photographic paper P; an exposure unit 24 that exposes the photographic paper P based on the print data; A development processing unit 25 including a plurality of processing tanks 25a, 25b, and 25c filled with processing solutions for developing, bleaching, and fixing the exposed photographic paper P is disposed along the transport path of the photographic paper P. The laterally-feeding conveyor 26 that discharges the photographic paper P that has been dried after the development process, and the sheet-paper (photo print) P that is stacked on the laterally-feeding conveyor 26 is sorted in order units. Configured to include the data 27.

  The exposure unit 24 receives a laser beam bundle of RGB three colors modulated based on the print data in the main scanning direction orthogonal to the conveyance direction with respect to the photographic paper P conveyed in the sub-scanning direction by the conveyance mechanism 28. An exposure head 24a for outputting and exposing is accommodated.

  A transport mechanism 28 including a plurality of roller pairs that transport the photographic printing paper P at a process speed corresponding to the exposure unit 24 and the development processing unit 25 disposed along the transport path is disposed before and after the exposure unit 24. Is provided with a chucker-type transport mechanism 28a capable of transporting photographic paper P in multiple rows.

  The controller 33 provided in the operation station 3 is installed with an application program that operates under the control of a general-purpose operating system and executes various controls of the photographic processing apparatus 1, and a monitor 34 as an operation interface with the operator. A keyboard 35, a mouse 36, and the like are connected.

  The photographic processing process executed by the cooperation of the hardware and software of the controller 33 will be described in functional blocks. As shown in FIG. 4, photographic image data read by the film scanner 31 and the media driver 32 is received. An image input unit 40 that performs predetermined preprocessing and transfers it to a memory 41, which will be described later, and print order information and image editing information are displayed on the screen of the monitor 34, and an operation for inputting necessary data for them. A graphic user interface unit 42 that generates a graphic operation screen for displaying an icon or generates various control commands based on an input operation from the keyboard 35 or mouse 36 to the displayed graphic operation screen, and the image input Photos transferred from the section 40 A memory 41 in which image data and photographic image data after correction processing by an image processing unit 47 (to be described later), correction parameters at that time, and set print order information are stored in a predetermined area, and print order information An order processing unit 43 for generating image data, an image processing unit 47 for performing density correction processing and contrast correction processing for each frame image or a predetermined number of frame images for each photographic image data stored in the memory 41, and A display control unit 46 including a video RAM for displaying the image data and various input / output graphic data developed in the memory 41 on the monitor 34 based on a display command from the graphic user interface unit 42; The final correction image after various correction processes is output to the photographic printer 2. A print data generating unit 44 for generating cement data, and a like formatter 45 which converts the final corrected image according to customer orders to a file format for writing in a storage medium such as a CD-R.

  The film scanner 31 is configured to operate in two modes: a pre-scan mode that reads an image recorded on the film F at a high speed although it is a low resolution, and a main scan mode that reads a high resolution at a low speed. Various correction processes are performed on the low-resolution image read in the can mode, and the final resolution for the high-resolution image read in the main scan mode based on the correction parameters stored in the memory 41 at that time. Correction processing is executed and output to the printer 2.

  When the film scanner 31 reads an image recorded on the film F, the film scanner 31 reads an unexposed portion between each frame image. The data of the unexposed portion is used for base density removal processing in the base density processing unit 10 of the color correction processing unit 471 described later.

  Similarly, the image file read from the media driver 32 includes a high-resolution captured image and its thumbnail image, and various correction processes described later are performed on the thumbnail image. Based on the stored correction parameters, a final correction process is performed on the high-resolution captured image. When the image file does not include a thumbnail image, the image input unit 40 generates a thumbnail image from the high-resolution captured image and transfers it to the memory 41. In this way, the calculation load of the controller 33 is reduced by executing various editing processes that are frequently trial and error on low-resolution images.

  The image processing unit 47 is configured to perform density correction processing, contrast correction processing, and the like for each frame image or a series of frame images for each photographic image data stored in the memory 41. However, in the color correction processing according to the present invention, the image processing unit 47 is configured to perform density correction processing, contrast correction processing, and the like for each frame image on each photographic image data stored in the memory 41. As shown in FIG. 1, a color correction processing unit 471 that performs color correction processing on the original image data, a density correction processing unit 472 that performs density correction processing on the original image data, A contrast correction unit 473 that performs contrast correction on the density correction curve generated for density correction by the density correction processing unit 472, a distortion correction unit 474 that corrects distortion caused by the photographing lens, and an edge of the image is enhanced. And a sharpening processing unit 475 for reducing noise and an enlargement / reduction processing unit 476 for converting the image size to a size suitable for a photographic print. That.

  The color correction processing unit 471 includes a base density processing unit 10 that performs base density removal processing on the original image data, and an average for obtaining an average pixel value for each color component for all the constituent pixels for each input frame image unit. A pixel value calculation unit 11; a coordinate conversion unit 12 that converts each pixel of the frame image into a pixel in a hue saturation coordinate system in which saturation is defined by a distance from the center and hue is defined by an angle around the center; A circular region that defines a plurality of concentric circular regions centered on the average pixel value in the hue saturation coordinate system and obtains a circular region characteristic value indicating a distribution characteristic of pixels included in each circular region for each region A characteristic value calculating means 13; a first correction amount calculating means 14 for obtaining a color correction amount based on the circular area characteristic value; and a color obtained by the first correction amount calculating means 14 in the hue saturation coordinate system. Expressed in correction amount A radial area characteristic value calculating means 15 for defining a plurality of radial areas in the circumferential direction at a predetermined angular interval around a pixel to obtain a radial area characteristic value indicating a distribution characteristic of pixels included in each radial area; The second correction amount calculating means 16 for obtaining a color correction amount based on the radial area characteristic value, and the color correction amount obtained by the first correction amount calculating means 14 or the second correction amount calculating means 16. The color correction unit 17 corrects the frame image based on the color correction amount.

The base density processing unit 10 is configured to perform a base density removal process on the original image data. Specifically, each of the base density processing units 10 read from an image recorded on the film F by the film scanner 31. When the density of the red component (R) in the unexposed portion of the frame image is r _b , the density of the green component (G) is g _b , and the density of the blue component (B) is b _b , each of the original image data adds _g b -r _b to R component of the pixel, it performs the process of adding a _g b -b _b B components. This treatment eliminates the effect of base density, which varies with film type.

  When the film scanner 31 cannot read the base density from the image recorded on the film F, or when using the image file read from the media driver 32, the base density is set for all the RGB components. Process as zero.

The average pixel value calculation means 11 is configured to obtain an average pixel value for each color component for all the constituent pixels for each inputted frame image. Specifically, the number of pixels the frame images is that an i number, n th RGB components respectively r _n of pixels of the frame _image, when g _n, and b _n, of each pixel of the frame image Average values avg (r), avg (g), and avg (b) of the RGB components are expressed as [Equation 1].

The coordinate conversion means 12 is configured to convert each pixel of the frame image into a pixel of a hue saturation coordinate system in which saturation is defined by a distance from the center and hue is defined by an angle around the center. cage, More specifically, into a xy coordinate system on the basis of the calculation expression shown each pixel r _n of the frame _image, g _n, the b _n in expression (2). The xy coordinate system is shown as a hue saturation coordinate system in which saturation is defined by a distance from the center and hue is defined by an angle around the center.

As shown in FIG. 5, when the positive direction of x in the xy coordinate system is set to 0 degree as shown in FIG. 5, the arithmetic expression shown in [Equation 2] is the direction of 0 degree and the G component is 120 degrees. And the B component as a direction of 240 degrees, and the size of each component as the distance from the origin in the direction of each component, conversion of the composite vector C _RGB of each component into the sizes of the x component and the y component Represents.

  The circular area characteristic value calculation means 13 defines a plurality of concentric circular areas centered on the average pixel value in the hue saturation coordinate system, and indicates a distribution characteristic of pixels included in each circular area The characteristic value is obtained for each region.

More specifically, as shown in FIG. 6, values obtained by converting the average pixel value into the xy coordinate system, that is, avg (r), avg (g), and avg (b) derived by [Expression 1] pixel _r n in _2], g n, plotted on the color saturation coordinate system _{b n x n} and _{y n} which is calculated by substituting the respective center coordinates avg (x) and as avg (y), the center coordinates Twelve concentric circular regions S1 to S12 are defined at a predetermined interval with respect to.

  In the circular area, the circular area located outside includes a circular area located inside the circular area. For example, the circular region S2 includes the circular region S1, and the circular region S12 includes the circular region S11 (that is, includes all of the circular regions S1 to S11). In this example, the number of circular areas is twelve, but the number of circular areas may be set as appropriate. Further, in this example, the predetermined interval between the circular regions is an equal interval, but the predetermined interval may be set as appropriate, and the size of the interval between the regions may be different evenly. They may be different from each other.

  The circular area characteristic value includes the number of pixels included in each circular area, and a difference value between the average pixel value and the average pixel value that is an average value for each color component of the pixels included in each circular area. Is done. The calculation of the circular area characteristic value is composed of four steps, which will be described in detail below.

In the first step, all the constituent pixels of the frame image are plotted in the hue saturation coordinate system, and the number of pixels m _Sj plotted for each of a plurality of regions Sj (j is an integer from 1 to 12) on the concentric circle is counted. To do.

In the second step, when the RGB components of the nth pixel among the pixels plotted in the region Sj are r _Sjn , g _Sjn , and b _Sjn , the average of the RGB components constituting the region Sj Pixel values avg (r _Sjn ), avg (g _Sjn ), and avg (b _Sjn ) are calculated based on the arithmetic expression shown in [Equation 3].

In the third step, the average pixel value, which is the average value of the RGB components of each pixel of the frame image, calculated by the average pixel value calculation means 11, and the average pixel value of each RGB component constituting the region Sj, Difference values Δr _Sj , Δg _Sj , and Δb _Sj are calculated based on the arithmetic expression shown in [Equation 4].

By normalizing the characteristic values m _Sj , Δr _Sj , Δg _Sj , Δb _Sj calculated from the above three steps to values in the range of 0 to 1, circular area characteristic values m _j , Δr _j , Δg _j , Δb _j is calculated. Specifically, the circular area characteristic value m _j is calculated by dividing the characteristic value m _Sj by the total number of pixels included in the frame image, and the circular area characteristic values Δr _j , Δg _j , Δb _j are The characteristic values Δr _Sj , Δg _Sj , Δb _Sj are respectively divided by the maximum density setting value in the frame image.

The characteristic values m _Sj , Δr _Sj , Δg _Sj , Δb _Sj may be used as the circular region characteristic values as they are.

As shown in FIG. 7, when a cumulative histogram of the number of pixels is shown with the region Sj as the horizontal axis and the number of pixels m _Sj of the pixels included in each circular region as the vertical axis, a frame image in which no color failure occurs is Since the distribution of data in the hue saturation coordinate system becomes small, the cumulative histogram tends to be saturated in the inner region (S5 in FIG. 7). Since the image has a large data distribution in the hue saturation coordinate system, the cumulative histogram saturates in an outer region (in FIG. 7, S7 is a different light source and S10 is a feria), that is, it is not easily saturated. There is a tendency.

The first correction amount calculation means 14 performs color correction amounts R1, G1 based on the circular area characteristic values m _j , Δr _j , Δg _j , Δb _j (or m _Sj , Δr _Sj , Δg _Sj , Δb _Sj ). , B1 is obtained. The color correction amounts R1, G1, and B1 are the differences between the original pixel values and the new pixel values, and the first correction amount calculation means 14 is, for example, a neural network that is learned by an error back propagation learning method, Alternatively, it is composed of multiple regression analysis means.

In the case of a neural network that is learned by the error back-propagation learning method, for example, as shown in FIG. 8, it is composed of a perceptron type composed of three layers: an input layer, an intermediate layer, and an output layer. An input layer to which circular region characteristic values m _j , Δr _j , Δg _j , Δb _j of each region Sj of the frame image are input, and an intermediate layer to which the result of weight calculation of input data with a predetermined connection weight is input The output of the intermediate layer is weighted with a predetermined coupling weight and is output as color correction amounts R1, G1, and B1, and is composed of three layers. Each coupling weight is preset by an error back propagation learning method. The

  Specifically, the neural network uses a color correction amount by manual correction performed on 9000 sample images with different shooting conditions and a circular area characteristic value for the sample images as teacher data, and each 100 million times for each sample image. The results are set so that repeated learning is performed and the result with the least variation is obtained.

When configured by multiple regression analysis means, for example, the circular region characteristic values m _j , Δr _j , Δg _j , Δb _j are respectively applied as dependent variables to the multiple regression equation as shown in [Equation 5]. Therefore, a configuration for deriving the color correction amounts R1, G1, and B1 for each of the RGB components as independent variables can be given.

In the above [Equation 5], A ₁₁ to A ₃₄ are coefficients, C ₁ to C ₃ are constant terms, and the relationship between the sample image and the correction amount when it is evaluated as appropriate by manual correction for a large number of sample images. It is a value calculated from

The radial area characteristic value calculation means 15 adds the color correction amounts R1, G1, and B1 obtained by the first correction amount calculation means 14 to the hue saturation coordinate system and the pixels r _n , g _n , and b _n , _{respectively.} A plurality of radial regions are defined in the circumferential direction at predetermined angular intervals with the pixel converted into the hue saturation coordinate system as a center, and a radial region characteristic value indicating a distribution characteristic of the pixels included in each radial region is defined for each region. It is configured to ask for.

More specifically, as shown in FIG. 9, by applying the color correction amount R1, G1, B1 derived in the first correction amount computing means 14, respectively pixel _{r n,} g n, the _{b n} in expression (2) the plotted in hue saturation coordinate system x _n and y _n calculated Te respectively center coordinates X, as Y, about said center coordinates, defining a plurality of radial regions in a circumferential direction at predetermined angular intervals. The number of radial regions is set as appropriate. For example, in FIG. 9, 12 radial regions are set every 30 degrees, and the following description will be made based on this setting.

  The radial region characteristic value is a weighted average pixel value for each color component obtained by multiplying the number of pixels included in each radial region by a larger weighting factor as the low saturation pixel is applied to the pixels included in each radial region. Composed. The calculation of the radial region characteristic value is composed of two steps, which will be described in detail below.

In the first step, all the constituent pixels of the frame image are plotted in the hue saturation coordinate system, and the number of pixels l _Hj plotted for each of a plurality of radial regions Hj (j is an integer from 1 to 12) is counted. .

In the second step, the RGB components of the nth pixel among the pixels plotted in the region Hj are _calculated as r _Hjn , g _Hjn , and b _Hjn based on the distance (saturation) from the center of each pixel. The radial region characteristic values Δr _Hj , Δg _Hj , Δb _Hj are calculated based on the arithmetic expression shown in [Equation 6], with the weighting factors as _{zr Hjn} , zg _Hjn , zb _Hjn , respectively. In [Equation 6], 1 added to the denominator and color correction amounts R1, G1, and B1 added to the numerator will be described later.

That is, a value obtained by multiplying each pixel existing in the region Hj by a weighting factor corresponding to each pixel is calculated, and the calculated values are summed and divided by the number of pixels existing in the region Hj, _lHj . Thus, the radial region characteristic values Δr _Hj , Δg _Hj , Δb _Hj are calculated. Further, the number of pixels l _Hj occupying the region Hj is also set as the radial region characteristic value. That is, the radial region characteristic values are four values of l _Hj , Δr _Hj , Δg _Hj , and Δb _Hj .

The weighting factors _{zr Hjn} , zg _Hjn , and zb _Hjn for each pixel (for example, the pixel values are r _Hjn , g _Hjn , and b _Hjn ) are normalized by the method described below for the density value of each pixel. It is a converted value. The normalization method normalizes the range of density setting values in the original image data (for example, the range of 0 to 255 when the original image data is composed of 8 bits) to the range of 0 to 4. Normalize by subtracting the normalized value from 1. If the normalized value becomes a negative value when subtracted from 1, the weight coefficient is set to 0. The weighting coefficient calculated in this way is increased in the weight of pixels with low saturation, and becomes 0 in the region corresponding to 75% on the high saturation side of the entire density range.

Further, in the calculation of [Equation 6], in addition to the pixels existing in the region Hj, the color correction amounts R1, G1, and B1 are added with the weighting factor being 1. As a result, the number of pixels increases by one, so in [ _Equation 6], the number of pixels is divided by l _Hj +1 instead of l _Hj . This is because there are pixels in the region Hj, but the weighting factor is 0 for all of the pixels, so the radial region characteristic value is 0, and there are no pixels in the region Hj. This is to distinguish that the radial region characteristic value is zero.

The second correction amount calculation means 16 obtains color correction amounts R2, G2, and B2 based on the radial region characteristic values l _Hj , Δr _Hj , Δg _Hj , Δb _Hj . The color correction amounts R2, G2, and B2 are the difference between the original pixel value and the new pixel value, and the second correction amount calculation unit 16 is, for example, similar to the first correction amount calculation unit 14, It consists of a neural network or multiple regression analysis means that is learned by the error back propagation learning method.

When the neural network is learned by the back propagation learning method, the radial region characteristic values l _Hj , Δr _Hj , Δg _Hj , Δb _Hj are input to the input layer shown in FIG. Color correction amounts R2, G2, and B2 for RGB components are output. Similarly to the above, the neural network also has a color correction amount by manual correction performed on 9000 sample images with different shooting conditions and a circular area characteristic value for each sample image as teacher data, and each 100 million each sample image. The result obtained by repeating the learning once and obtaining the result with the least variation is set.

When configured by multiple regression analysis means, for example, the radial region characteristic values l _Hj , Δr _Hj , Δg _Hj , Δb _Hj are applied as dependent variables to the multiple regression equation as shown in [Equation 5], respectively. Therefore, a configuration for deriving the color correction amounts R2, G2, and B2 for each of the RGB components as independent variables can be given.

  The color correction unit 17 is configured to correct the frame image based on the color correction amount obtained by the second correction amount calculation unit 16, and more specifically, for all pixels of the frame image, The frame image is corrected by adding the color correction amounts R2, G2, and B2 obtained by the second correction amount calculator 16.

  When correcting the frame image according to the present invention described above, when the correct pixel value is a pixel value evaluated as appropriate by manual correction, the probability of obtaining the correct pixel value depending on the degree of key correction and an incorrect answer As shown in FIG. 10, the standard deviation of the error is a method using conventional LATD correction, a method using the first correction amount calculating unit 14 and not using the second correction amount calculating unit 16, and A method using the second correction amount calculation means 16 in addition to the first correction amount calculation means 14 will be described.

  In FIG. 10, the method using the first correction amount calculating means 14 is improved over the conventional method using LATD correction, and the method using the second correction amount calculating means 16 is the first correction amount calculating means. This is an improvement over the method using the correction amount calculation means 14.

  In the photographic image processing apparatus according to the present invention described above, the average pixel value calculation means 11 for obtaining the average pixel value for each color component for all the constituent pixels for each inputted frame image unit, and each pixel of the frame image from the center. A coordinate conversion means 12 for converting into pixels of a hue saturation coordinate system in which saturation is defined by a distance and a hue is defined by an angle around the center, and a concentric circle centered on the average pixel value in the hue saturation coordinate system A circular area characteristic value calculating means 13 for defining a circular area characteristic value indicating a distribution characteristic of a pixel included in each circular area, and color correction based on the circular area characteristic value An image processing program for causing a computer to function as the first correction amount calculation unit 14 for determining the amount, and the hue saturation coordinate system are represented by the color correction amount obtained by the first correction amount calculation unit 14. A plurality of radial regions in the circumferential direction at a predetermined angular interval around the pixel, and a radial region characteristic value calculating means 15 for obtaining a radial region characteristic value indicating a distribution characteristic of pixels included in each radial region for each region; This is easily realized by installing an image processing program for causing the computer to function as the second correction amount calculation means 16 for obtaining the color correction amount based on the radial region characteristic value.

Hereinafter, another embodiment will be described. In the above-described embodiment, the radial region characteristic value calculation unit 15 uses the color correction amounts R1, G1, and B1 derived by the first correction amount calculation unit 14 as pixels r _n , g _n , and b _n , _respectively , [Equation 2]. In the above description, x _n and y _n calculated by applying to the hue saturation coordinate system are plotted as center coordinates X and Y, respectively. However, the center coordinates are avg (x) and avg (y) shown in FIG. ).

That is, the central coordinates are values obtained by converting the average pixel value calculated by the average pixel value calculating means 11 into the xy coordinate system, that is, avg (r), avg (g), avg (b ) And x _n and y _n calculated by substituting the pixels r _n , g _n and b _n in [Equation 2], _respectively , as center coordinates avg (x) and avg (y) are plotted in the hue saturation coordinate system. To do.

  In the above-described embodiment, the first correction amount calculation unit 14 and the second correction amount calculation unit 16 are configured by a neural network or a multiple regression analysis unit that is learned by an error back propagation learning method. However, other than the neural network and the multiple regression analysis means, for example, a configuration using a lookup table may be used.

More specifically, in the case of a look-up table, the first correction amount calculation means 14 uses circular region characteristic values m _j , Δr _j , Δg _j , Δb _j according to a predetermined function stored in advance. Is converted into color correction amounts R1, G1, and B1. In the second correction amount calculation means 16, the radial area characteristic values l _Hj , Δr _Hj , Δg _Hj according to a predetermined function stored in advance. , Δb _Hj are converted into color correction amounts R2, G2, and B2.

In the above-described embodiment, the conversion to the hue saturation coordinate system has been described for the configuration in which each pixel of the frame image is converted to the xy coordinate system based on the arithmetic expression shown in [Expression 2]. The conversion to is not limited to the conversion based on the arithmetic expression shown in [Expression 2]. For example, each pixel r _n of the frame _image, g _n, a configuration in which the horizontal axis based on the matrix equation shown in the b _n [Math 7] is P, the vertical axis is converted to the hue saturation coordinate system represented by Q There may be.

  In the above-described embodiment, a description has been given of a configuration in which a greater weighting factor is applied to a pixel included in each radial region in a calculation of the radial region characteristic value. However, a criterion for the low saturation pixel is appropriately set. However, the present invention is not limited to this embodiment.

  In the above-described embodiment, the circular region characteristic value includes the number of pixels included in each circular region, the region average pixel value that is an average value for each color component of the pixels included in each circular region, and the average pixel value. However, the circular area characteristic value may be the area average pixel value.

  In the above-described embodiment, the circular region characteristic value includes the number of pixels included in each circular region, the region average pixel value that is an average value for each color component of the pixels included in each circular region, and the average pixel value. The circular region characteristic value is the number of pixels included in each circular region and the average value for each color component of the pixels included in each circular region. An area average pixel value can also be adopted.

In this case, when the first correction amount calculation means 14 is constituted by a neural network learned by the error back propagation learning method, for example, as shown in FIG. 11, three layers of an input layer, an intermediate layer, and an output layer The average value avg (r), avg (g), avg (b) of the RGB components of each pixel obtained by the average pixel value calculation means is a value avgn (normalized). r), avgn (g), avgn (b), and the circular region characteristic values m _j , avg (r _jn ), avg (g _jn ), avg (b _jn ) of each region Sj (these are expressed by [ _Equation 3] An input layer to which average pixel values avg (r _Sjn ), avg (g _Sjn ), and avg (b _Sjn ) are respectively input), and a result of weight calculation of the input data with a predetermined connection weight Enter And the output of the intermediate layer is output as color correction amounts R1, G1, and B1 by weight calculation with a predetermined connection weight, and each connection weight is determined by an error back propagation learning method. It is set in advance.

  In the above-described embodiment, when the first correction amount calculation unit 14 and the second correction amount calculation unit 16 are configured by a neural network, the intermediate layer of the neural network is configured by one layer. The intermediate layer may be composed of a plurality of layers.

  In the above-described embodiment, the color correction unit 17 corrects the frame image based on the color correction amount obtained by the second correction amount calculation unit 16, but the second correction amount calculation unit has been described. Even if the frame image is corrected based on the color correction amount obtained by the first correction amount calculation means 14 without providing 16, the predetermined effect can be obtained.

  Note that each of the above-described embodiments is merely an example of the present invention, and a plurality of embodiments can be combined within the scope of the effects of the present invention, and the specific configuration of each block is appropriately set. It is also possible to design changes.

Functional block diagram of the image processing unit Explanatory drawing of the external configuration of the photographic image processing apparatus Illustration of photo printer Explanatory drawing of functional block configuration of photographic image processing apparatus Illustration of hue saturation coordinate system Illustration of multiple concentric regions in the hue saturation coordinate system Explanatory drawing of the cumulative histogram of the number of pixels in each concentric area Explanatory drawing of the first correction amount calculation means composed of a neural network Illustration of multiple radial regions in the hue saturation coordinate system Explanation table of accuracy rate of color correction according to the present invention Explanatory drawing of the 1st correction amount calculating means comprised with the neural network which shows another embodiment

10: Base density processing unit 11: Average pixel value calculation means 12: Coordinate conversion means 13: Circle area characteristic value calculation means 14: First correction amount calculation means 15: Radial area characteristic value calculation means 16: Second correction amount calculation means 17: Color correction means

Claims (8)

Mean pixel value calculation means for obtaining an average pixel value for each color component for all constituent pixels for each inputted frame image, and saturation of each pixel of the frame image is defined by a distance from the center at an angle around the center Coordinate conversion means for converting into pixels in a hue saturation coordinate system in which hue is defined, and in the hue saturation coordinate system, pixels represented by average pixel values for each color component are converted into the hue saturation coordinate system A circular region characteristic value calculating means for defining a plurality of concentric circular regions centered on the selected pixel and obtaining a circular region characteristic value indicating a distribution characteristic of the pixels included in each circular region for each region; and the circular region characteristic A photographic image processing apparatus comprising first correction amount calculation means for obtaining a color correction amount based on a value. The circular area characteristic value includes the number of pixels included in each circular area and the average value of each pixel color component included in each circular area, or the area average pixel value and the color The photographic image processing apparatus according to claim 1, wherein the photographic image processing apparatus includes a difference value from an average value of average pixel values for each component .   3. The photographic image processing apparatus according to claim 1, wherein the first correction amount calculation means is configured by a neural network learned by an error back propagation learning method or multiple regression analysis means. In the hue / saturation coordinate system, the color correction amount derived by the first correction amount calculation unit is converted into the hue / saturation coordinate system as pixels, and the converted pixels are centered in the circumferential direction at predetermined angular intervals. A plurality of radial areas are defined, a radial area characteristic value calculating means for obtaining a radial area characteristic value indicating the distribution characteristic of the pixels included in each radial area for each area, and a color correction amount is obtained based on the radial area characteristic value. The photographic image processing apparatus according to claim 1, further comprising second correction amount calculation means.   The radial region characteristic value is a weighted average pixel value for each color component obtained by multiplying the number of pixels included in each radial region by a larger weighting factor as the low saturation pixel is applied to the pixels included in each radial region. The photographic image processing apparatus according to claim 4 configured.   6. The photographic image processing apparatus according to claim 4, wherein the second correction amount calculation means is configured by a neural network learned by an error back propagation learning method or multiple regression analysis means. An average pixel value calculation step for obtaining an average pixel value for each color component with respect to all constituent pixels for each inputted frame image, and for each pixel of the frame image, saturation is defined by a distance from the center and an angle around the center. A coordinate conversion step for converting into pixels in a hue saturation coordinate system in which hue is defined, and a pixel represented by an average pixel value for each color component is converted into the hue saturation coordinate system in the hue saturation coordinate system A circular region characteristic value calculating step for defining a plurality of concentric circular regions centered on the selected pixel and obtaining a circular region characteristic value indicating a distribution characteristic of the pixels included in each circular region for each region; and the circular region characteristic And a first correction amount calculation step for obtaining a color correction amount based on the value. In the hue / saturation coordinate system, the color correction amount derived in the first correction amount calculation step is converted into the hue / saturation coordinate system as pixels, and the converted pixels are centered in the circumferential direction at predetermined angular intervals. A plurality of radial areas are defined, a radial area characteristic value calculation step for obtaining a radial area characteristic value indicating the distribution characteristics of pixels included in each radial area for each area, and a color correction amount is obtained based on the radial area characteristic value. The photographic image processing method according to claim 7, comprising a second correction amount calculation step.

Images