JP2008072551A - Image processing method, image processing apparatus, program and recording medium - Google Patents

Abstract

<P>PROBLEM TO BE SOLVED: To achieve dynamic range compression while reducing operation load and utilizing color reproduction capability of an output device in converting scene reference image data into image data for the output device. <P>SOLUTION: An image analyzing section 103 analyzes color distribution of scRGB image data to set range compression conditions for tone mapping processing. A range compressing section 101 compresses a dynamic range on the basis of the range compression conditions set in the image analyzing section 103, and converts the scRGB image data into data within a range expressible in an expanded RGB color space. A color correction section 102 converts the color signals compressed into the dynamic range of the expanded RGB color space into CMYK signals as printer output signals. <P>COPYRIGHT: (C)2008,JPO&INPIT

Description

  The present invention relates to an image processing technique for generating preferable output image data for image data with a wide dynamic range, and in particular, an image processing method for performing image conversion for outputting image data to an output device such as a monitor or a printer, The present invention relates to an image processing apparatus, a program, and a recording medium.

  In recent years, with the widespread use of personal computers, the Internet, and home printers, and the increase in storage capacity of hard disks and the like, opportunities to handle photographic images with digital data have increased, and as a means of inputting digital photographic image data (hereinafter referred to as image data). Digital cameras are popular.

  An image sensor CCD (Charge Coupled Device) used in the digital camera outputs image data that is substantially proportional to the luminance of the shooting scene. If the imaging data from the CCD is output to the monitor as it is, an image having a luminance proportional to the power of the luminance is displayed, which is not a preferable characteristic. For this reason, it is necessary to perform image conversion in accordance with the characteristics of the output device. In general, a standardized color space called sRGB based on an average monitor characteristic (see Non-Patent Document 1) is widespread, and digital cameras are often converted into sRGB image data and output.

  As described above, in the conventional workflow, by using sRGB as the standard color space, it is possible to obtain a generally good color reproduction even when displayed on the monitor as it is.

  By the way, since sRGB is a color space matched to an average monitor as described above, the color reproduction range is limited to a range substantially similar to that of the monitor. On the other hand, the color reproduction range of the printer has a part wider than the monitor. For example, there is a region that cannot be reproduced by sRGB in the cyan region of an inkjet printer. In addition, there are colors belonging to these areas in the subject to be photographed, but such colors are compressed into the sRGB color reproduction range even though the printer has the ability to reproduce them. It becomes impossible to reproduce.

  Furthermore, there are various shooting scenes in the natural world, and the luminance range of an actual scene often occurs on the order of several thousand: 1 outdoors. Therefore, various methods have been proposed for expanding the dynamic range of digital cameras, for example, by combining images using two types of CCDs having different sensitivities. However, since the sRGB space is limited to the luminance range of the monitor, the dynamic range is insufficient to represent an actual scene.

  Therefore, recently, a technique has been proposed in which image data associated with characteristics of an actual shooting scene is output from a digital camera instead of a color space adapted to a monitor such as sRGB. In other words, image data proportional to the brightness of the actual shooting scene, converted to a color space defined calorimetrically, is output from the digital camera, or it is intentionally corrected even if it is not image data proportional to the brightness. -Output image data without any emphasis. A color space that can be converted into scene luminance / chromaticity values by conversion that can be described by a simple mathematical expression is called a scene reference color space. For example, RIMM RGB or scRGB (IEC standard 61966-2-2) is known as the scene reference color space. Also, image data represented by color signals in the scene reference color space is called scene reference image data. For example, Raw image data generally used in a digital camera is also subjected to matrix calculation representing the characteristics of the image sensor. Thus, since it can be converted into the luminance value of the scene, it can be regarded as scene reference image data. On the other hand, image data matched with a conventional output device such as sRGB is called output reference image data.

  By using the above-mentioned scene reference color space, a dynamic range wider than that of the sRGB color space can be expressed. Therefore, image information with no gradation collapse can be obtained for various shooting scenes from a bright scene in the daytime to a dark scene such as a night view. Can be recorded. However, since the dynamic range of monitors and printers is significantly narrower than actual scenes, dynamic range compression suitable for image data is required to reproduce images on output devices while taking advantage of the scene reference color space. become.

  The following methods have been proposed as a method for performing range compression on image data having a wide dynamic range as described above.

  Determine whether the input image data is within the range of 0 to 100% or not, analyze the image data outside the range, set the white reference point or black reference point, and the image data below 100% is high gradation A device that performs color conversion with a DLUT and performs color conversion with a low-tone DLUT when RGB values exceeding 100% are included (see Patent Document 1), and scene types such as backlight and high contrast using a scene reference image After determining and setting a white point or black point in a specific scene, tone correction is performed so that the main subject has an appropriate brightness, and the color correction model is used for the image data after tone correction. A device that converts image data suitable for the output device (see Patent Document 2), calculates appearance parameters based on EXIF information, analyzes scene reference image data, and includes a person in the scene Modify the appearance parameters for determining whether there is a device for converting the image data visible Preferably (see Patent Document 3) on the basis of the parameters.

Multimedia Systems and Equipment-Color Measurement and Management-Part2-1: Color Management-Default RGB Color Space-sRGB IEC 61966-2-1 CIE TECHNICICAL REPORT-CIE159: 2004 JP 2006-80834 A Japanese Patent Application Laid-Open No. 2005-209012 Japanese Patent Laying-Open No. 2005-210526 JP 2002-368893 A

  When scene reference image data represented by scRGB is output from a digital camera, it is necessary to convert it into image data for an output device. However, in the conventional method, the difference in visual characteristics due to the difference in the observation conditions between the shooting scene and the output device cannot be sufficiently corrected, so that the scene reference image data cannot be converted into image data for a preferred output device.

  In the above-described method of Patent Document 1, scRGB image data is analyzed and a DLUT for converting an input RGB signal into an output CMYK signal is adjusted based on a white reference value or a black reference value, and the input RGB signal and the output CMYK signal are strong. Since it has non-linear characteristics, a three-dimensional lookup table is used as the DLUT. For this reason, in order to automatically perform the process of adjusting the DLUT in accordance with the change of the white reference value or the black reference value, a complicated calculation is required, which is difficult to realize. Also, regarding the method of determining the white reference value, the highest luminance value in the image data is simply detected, but if the light source color is included in the image data, the highest luminance value is determined as the white reference value. Then, it often becomes dark in the main subject area.

  In the dynamic range compression method of Patent Document 2, tone correction is performed using the reference white point and the reference black point set by analyzing the scRGB image so as to fall within the dynamic range of the sRGB space. Since the correction is compressed to the gradation range of sRGB, for example, cyan that can be reproduced by the printer is also compressed within the sRGB color reproduction range, and the merit of the scRGB image is lost. In addition, since uniform gradation correction is performed, there is a problem that the brightness of the subject and the gradation reproduction of the highlight are not compatible.

  Further, even in the method of setting the parameters of the color appearance model according to Patent Document 3 in accordance with the scene, it is necessary to perform a complicated color appearance calculation for each image in order to achieve appropriate gradation characteristics according to the image. The calculation load is large.

The present invention has been made in view of the above problems,
SUMMARY OF THE INVENTION An object of the present invention is to provide an image processing method and image processing that realizes dynamic range compression utilizing a color reproduction capability of an output device with a small calculation load when converting scene reference image data to image data for an output device. To provide an apparatus, a program, and a recording medium.

Another object of the present invention is to provide an image processing method, an image processing apparatus, a program, and a recording medium that perform dynamic range compression that can achieve both of the brightness of a subject and gradation reproduction of highlights.
That is,
An object of claim 1 is an image that realizes dynamic range compression that uses less computational load and takes advantage of the color reproduction ability of an output device when converting scene reference image data having a wide dynamic range into image data for an output device. It is to provide a processing method.

  An object of a second aspect of the present invention is to provide an image processing method of the first aspect of the present invention, which has a low computational load and can perform dynamic range conversion only by calculation for each color component of scene reference image data.

  An object of the present invention is to provide an image processing method for performing preferable dynamic range conversion according to the characteristics of scene reference image data.

  An object of the present invention is to provide a preferable dynamic range conversion method for a photographic image not including a person.

  An object of the present invention is to provide a preferable dynamic range conversion method for an image in which the luminance level of a person portion as a subject is appropriate.

  An object of the present invention is to provide a preferable dynamic range conversion method for a backlight image or a high-contrast image in which the luminance level of the person portion that is the subject is inappropriate.

  An object of a seventh aspect of the present invention is to provide an image processing method according to the sixth aspect of the present invention, which can suppress a gradation jump at the boundary between the subject area and the non-subject area.

  The purpose of claim 8 is to convert scene reference image data having a wide dynamic range into ornamental image data that can be seen favorably on the output device, and to reduce dynamic load and to take advantage of the color reproduction capability of the output device. An object of the present invention is to provide an image device that realizes the above.

  The object of the present invention is to reduce dynamic load when converting scene reference image data having a wide dynamic range into ornamental image data that can be seen favorably on the output device, and to take advantage of the color reproduction capability of the output device. It is to provide a color conversion program for realizing the above.

  The present invention is an image processing method for converting an input color signal in a scene reference color space into a color signal for an output device, analyzing a color distribution of the input color signal, and performing gradation based on the analysis result Compression conditions are determined, and the input color signal is mapped to an extended color space that is narrower than the scene reference color space and includes the color reproduction range of the output device according to the gradation compression conditions, and is mapped to the extended color space The most important feature is color conversion of a color signal into a color signal for an output device.

  According to the present invention (claims 1, 8, and 9), gradation compression conditions suitable for the color distribution of the input image data are determined, and the color signal in the scene reference color space is included in the color reproduction range of the output device. Since gradation compression is performed in the extended color space, it is possible to perform range compression of a high dynamic range image using 100% of the color reproducibility of the output device with a simple calculation.

  According to the present invention (Claim 2), since the color space that can be converted from the color signal represented in the scene reference color space using a one-dimensional lookup table is used as the extended color space, the calculation load is reduced. it can.

  According to the present invention (Claim 3), since the luminance of the subject region and the luminance range of the image are extracted at the time of analyzing the color distribution, it is possible to perform dynamic range compression according to the characteristics of the image.

  According to the present invention (Claim 4), when the scene reference image does not include a person region, the range of the input signal is compressed by performing linear luminance conversion, and the color in the extended color space is converted from the compressed color signal. Since it is converted into a signal, complicated non-linear processing can be made common regardless of the image, and calculation processing for range correction can be simplified.

  According to the present invention (Claim 5), when the luminance of the human area of the scene reference image is appropriate, linear luminance conversion is performed to compress the range of the input signal, and the compressed color signal is converted into the extended color space. Therefore, complicated non-linear processing can be made common regardless of an image, and range correction calculation processing can be simplified.

  According to the present invention (Claim 6), when the luminance of the human area of the scene reference image is inappropriate, different range compression is performed for the human area and the non-human area, so the brightness of the subject and the gradation of highlight Both reproduction can be achieved.

  According to the present invention (Claim 7), in the invention of Claim 6, since the conversion characteristic is set using the luminance distribution characteristic near the target pixel and the attribute of the target pixel, the boundary between the subject area and the non-subject area Therefore, it is possible to suppress the generation of an unnatural pseudo contour.

  Hereinafter, embodiments of the present invention will be described in detail with reference to the drawings.

Example 1:
1. Image Processing System FIG. 3 shows a configuration example of the image processing system of the present invention. In FIG. 3, reference numeral 203 denotes a computer on which various color correction applications and software such as a printer driver are mounted. A digital camera 201 and a scanner 202 are input devices for capturing image data to be processed. The display 200 is an output device for displaying image data, and the color printer 204 is an output device for printing out image data. The color printer may be a color copier or a color facsimile machine.

  In the image processing system of FIG. 3, a digital camera 201 and a scanner 202 output image data representing a photograph or the like as a matrix pixel to the computer 203. The computer 203 converts the input image data into image data suitable for display display or print output, and displays the converted RGB image data on the display 200, or CMY or CMYK 2 via a printer driver. It has a function of converting it into value data and causing the color printer 204 to print it. The display 200 and the printer 204 display / output the image processed image data as a dot matrix image.

2. High Dynamic Range Image Data Next, image data input by the imaging device of the digital camera 201 and the scanner 202 will be described. In a conventional digital camera, it is common to convert image data into sRGB data suitable for display display in the digital camera and convert the format to EXIF format, and then send the image data to a computer. However, various scene reference color spaces have been proposed because the color reproduction range of the sRGB color space is narrow. The scRGB color space (IEC standard 61966-2-2) has been attracting particular attention in recent years because it will be adopted in the next Windows system.

  In the following description, the scRGB image is used as the scene reference image data in the present embodiment, but scene reference image data represented in other color spaces such as RAW data may be used.

  The scRGB image data will be described. According to the IEC standard 61966-2-2, the relationship between the scRGB data and the XYZ tristimulus values is defined as follows.

  As can be seen from the above definition, the scRGB signal is obtained from the XYZ tristimulus values by linear conversion. The luminance range of Y is defined as -0.5 to 7.5, and image data including white lighter than the reference white point (Y = 1.0) can be handled. In addition, since the conventional sRGB signal is 8-bit data (0 to 255), the scRGB signal is 16-bit data (0 to 65535), so that more gradations can be expressed than sRGB. The wide dynamic range information of the actual shooting scene can be expressed without loss. An image that reproduces a dynamic range wider than the conventional one is referred to as a high dynamic range image (HDR image). For example, in the case of an image in which bright light is inserted into the background of a person as shown in FIG. 2A, when the camera performs exposure control with focus on the person and generates an sRGB image, the gradation of the background is lost. In the example of the luminance histogram in FIG. 2B, it can be seen that the frequency at the point representing white (the right end in the figure) is remarkably high, and the gradation is crushed by the highlight. On the other hand, in the case of scRGB image data, a luminance exceeding 1 can be expressed, so that the backlit portion of the background can be output as image data with no gradation collapse.

3. Overall Flow of Image Processing FIG. 1 shows a configuration of an image processing apparatus provided in the computer 203. The image processing apparatus includes a range compression unit 101, a color correction unit 102, and an image analysis unit 103. The image analysis unit 103 analyzes the color distribution of the scRGB image data and sets a range compression condition for tone mapping processing. The range compression unit 101 performs dynamic range compression based on the range compression conditions set by the image analysis unit 103. Here, the range compression unit 101 converts the scRGB image data into a range range that can be expressed in an extended RGB color space described later. The color correction unit 103 performs a memory map interpolation operation on the color signal compressed to the extended RGB dynamic range, and converts it into a CMYK signal that is a printer output signal.

  The relationship of the color space reproduction range in the conversion will be described with reference to FIG. The color reproduction range of the scRGB color space, which is the input color space, is very wide and can express a dynamic range of -0.5 to 7.5. On the other hand, the sRGB space represents a standard display color reproduction range, and cannot cover the printer color reproduction range as shown in FIG. Therefore, if the scRGB image is converted into the sRGB image range, the color reproduction range of the printer cannot be used 100%.

  Therefore, in the present invention, the color reproduction capability of the printer is made 100% effective by performing range compression to the extended RGB space as a color space that can cover the reproduction range of the printer and then converting the color to a printer signal. The extended RGB color space is narrower than the scRGB color space, but is a color space that includes both the sRGB color space and the printer color space. As such a color space, the AdobeRGB space is known. However, in order to convert the scRGB color signal into the AdobeRGB signal, a matrix operation is required, and the calculation is complicated. Therefore, in this embodiment, an e-sRGB color space that can be converted from the scRGB space by simple table conversion is used. The relationship between e-sRGB and XYZ tristimulus is defined as follows.

  When the number of bits of e-sRGB is n and the component value of the e-sRGB signal is C, normalization is performed by Expression (3). The number of bits of the e-sRGB signal is standardized to 10 bits, 12 bits, and 16 bits.

Next, in accordance with the value of C ′, nonlinear conversion is performed to convert it into a luminance linear signal.
(1) When C ′ <− 0.04045

(2) When -0.04045 <= C '<= 0.04045

(3) When C ′> 0.04045

そして、下記のマトリックス変換でＸＹＺ三刺激値に変換できる。

And it can convert into an XYZ tristimulus value by the following matrix conversion.

  In the present embodiment, the number of bits of the e-sRGB signal is described as 10 bits in consideration of the gradation reproduction capability of the printer, but is not limited thereto. Further, as can be seen from the above equation, the matrix conversion coefficients are the same as scRGB and sRGB. Therefore, as for the relationship between e-sRGB and scRGB, the chromaticity values of the three primary colors and the white point are the same, and only their luminance ranges are different. The luminance level = 0 is 384 for e-sRGB and 4096 for scRGB. The luminance level = 1 is 894 for e-sRGB and 12288 for scRGB. FIG. 5 shows the characteristics of the scRGB signal and the e-sRGB signal.

4). 4. Detailed Description 4.1 Operation Description of Image Analysis Unit 103 The operation of the image analysis unit 103 will be described with reference to the flowchart of FIG. As a range compression method, a method of linearly compressing the luminance so that the maximum luminance of the input image matches the maximum luminance value of the e-sRGB image after the range compression can be considered. However, in a scene with a very strong light in the surroundings, such as a snowy scene or a summer beach, the subject person often appears relatively dark. The same applies to a backlight image, and the face of a person becomes dark when flash photography is not performed. Thus, even if linear range compression processing is performed on an image including a region with a high luminance level in the background portion of the shooting scene, it is impossible to achieve both the brightness of the subject and the gradation reproduction of the background.
Therefore, in this embodiment, when compressing the dynamic range, the range compression method is controlled based on the luminance level of the subject area of the image and the luminance level of the background area.
First, in step S1, the range of the input image is calculated. Specifically, a cumulative frequency distribution is generated for each of r, g, and b components of scRGB data, and a component value corresponding to a cumulative frequency of 99% is obtained. The three maximum component values obtained are set as maxR, maxG, and maxB, respectively. The maximum value of these three maximum component values is regarded as the range of the input image and is set to maxC. That is, in the range compression process, the scRGB signal is processed so that the range of 0 to maxC falls within the range of e-sRGB.

If maxC is less than scRGB signal value 12288 corresponding to luminance 1.0, the luminance range is expanded by performing the following conversion so that maxC becomes 12288. That is, Rsc ′ = (8192 × (Rsc−4096) / (maxC−4096)) + 4096
Gsc ′ = (8192 × (Gsc−4096) / (maxC−4096)) + 4096
Bsc ′ = (8192 × (Bsc−4096) / (maxC−4096)) + 4096
In the above, the luminance range of scRGB is in the range of -0.5 to 7.5. In order to save memory, a cumulative frequency distribution with 256 gradations is first created, and the cumulative frequency distribution is 99%. The luminance value Y1 that is less than the maximum luminance Y2 that is 100% may be obtained, and the cumulative distribution may be obtained by further subdividing between Y1 and Y2.

  Next, in step S2, the white point of the input image is obtained. The extraction of the white region is based on the luminance histogram, and when there is no average value of 95% or more of the maximum luminance and a saturation value of 10 or less and a color with a saturation value of 10 or less does not exist, WhitePoint does not exist. Use the default value.

Next, the process proceeds to step S3, and the face area is detected. In recent years, as a face detection method, a pattern matching method or a detection method using a learning model such as a neural network has been proposed and has already been implemented in products. If the face area is detected, the average luminance SkinY is next obtained for the pixels included in the skin color hue of the face area (step S4).
However, if the image does not include a face, SkinY = −1.

  The range compression processing is performed by sending maxC, WhiteY, and SkinY obtained by the above processing to the range compression unit 101.

4.2 Operation of Range Compression Unit 101 Next, the range compression unit 101 performs scRGB signal luminance conversion according to the subject luminance SkinY and white point WhiteY obtained by the image analysis unit 103 and the range maxC of the input image, and then e. -Perform nonlinear conversion to convert to sRGB signals. Hereinafter, the range compression process will be described with reference to the flowchart of FIG.

  In step S12, it is determined whether SkinY = −1. When SkinY is -1, it means that no subject is included in the image. In such a case, linear luminance range conversion is performed in step S14 with reference to the input image ranges minC and maxC. Next, in step S13, it is determined whether or not the subject brightness SkinY is appropriate. As a determination method, the average luminance of the subject is compared with the dynamic range of the image, and it is determined whether the luminance of the subject is appropriate for the dynamic range. For example, if the ratio of SkinY / maxC is within a predetermined range, it is regarded as appropriate. The appropriate ratio is set in advance by performing a sensory evaluation experiment. If the ratio is appropriate, the person is corrected to a preferable brightness by linearly compressing the luminance as described above.

For an image that does not include a person or an image with an appropriate person brightness level, a linear range compression process is performed in step S14. The range processing conversion method will be described below.
(1) Luminance conversion Based on the obtained reference white point maxC, the luminance of the scRGB value is converted. Since scRGB is originally a luminance linear signal, the hue is maintained if the ratio of each of the r, g, and b components is constant. The luminance conversion can be performed by the following primary conversion.

scd = (sc−4096) * (17694-4096) / (maxC−4096) +4096
Here, sc is an input value of the scRGB signal, and scd is a signal value after luminance conversion. With the above calculation, the input data is compressed within the range of the e-sRGB signal. FIG. 8 shows an example of input / output characteristics by range correction.
(2) scRGB → e-sRGB conversion The scRGB signal compressed within the range of e-sRGB is converted to esRGB. Since the scRGB signal is a 16-bit signal, it can be converted by a one-dimensional lookup table, but the 16-bit lookup table consumes a large amount of memory. Therefore, it is converted into an e-sRGB signal using a conversion formula. That is,
x = (scd / 8192.0) −0.5
if (x <−0.0031308)
x ′ = − (1.055 × (−x) ^ (1 / 2.4) −0.055)
if (−0.0031308 <= x <0.0031308)
x ′ = 12.92 × (x)
if (x> 0.0031308)
x ′ = 1.055 × (x) ^ (1 / 2.4) −0.055
esd = 255.0 * 2 ^ (n-9) * x '+ offset
Can be converted. Here, esd is an e-sRGB signal value, and offset is 384 in the case of a 10-bit e-sRGB signal. The conversion characteristics of the above conversion are shown in FIG.

  If it is determined in step S13 that the subject brightness is not appropriate, the process proceeds to step S15, where the entire image is analyzed and separated into a background region and a subject region. As a background / subject region separation method, for example, the method disclosed in Patent Document 4 can be used. The separation result is recorded in a 1-bit layer memory in which the subject area is 0 and the background area is 1. FIG. 10 shows an example of the separation result.

  When the background area and the subject area are separated, a range correction table is created in step S16 in order to obtain an image with good gradation reproducibility in both the background area and the subject area. At this time, for the subject area, range correction determined based on the subject is performed, and in the highlight area including high-intensity white light, the range correction is dynamically performed in the image so as to perform range compression suitable for the highlight area. Change the table. For example, the correction table of FIG. 11A is used for the subject area, and the correction table of FIG. 11B is used for the highlight area. However, as described above, in order to save memory, it is better to perform conversion using an arithmetic expression than a one-dimensional lookup table.

  Here, continuous conversion is necessary so that a pseudo contour does not occur at the boundary between the subject region and the background region. Therefore, a correction table is created with reference to surrounding N × N pixels centered on the target pixel. First, the ratio Rs between the subject area and the background area in N × N pixels is obtained. Next, the luminance range in N × N pixels is obtained. When the luminance range is small, a correction table for the subject area is used for the subject area, and a correction table for the background area is used for the background area. When the luminance range is greater than or equal to a predetermined value, the correction amount is changed according to the value of Rs. The correction amount can be changed by using a method in which an inner product operation is performed using Rs for the result of subject correction and the result of background region correction.

Once the range correction table is created, range correction is performed in step S17. For example, when the range correction table of FIG. 8 is set, if the scRGB image data is (inr, ing, inb), the maximum value is inY and it is converted to outY. If the maximum value is G component, ing = inY,
inr ′ = inr−4096
ing '= ing-4096
inb ′ = inb−4096
outr = inr ′ × (outY−4096) / (inY−4096) +4096
outg = outY
outb = inb ′ × (outY−4096) / (inY−4096) +4096
The color is not changed before and after conversion. Then, the conversion to the e-sRGB signal is performed using the above-described nonlinear conversion formula from the scRGB signal to the e-sRGB signal.

4.3 Color Correction Unit 102
Next, the color correction unit 102 converts the image data converted by the range compression unit 101 into a printer signal. The color correction unit 102 converts the color signal compressed into the e-sRGB space by the range compression unit into a printer color signal by performing an interpolation operation using a three-dimensional lookup table. The e-sRGB space includes a printer color space and has a wider dynamic range than a hard copy. That is, the dynamic range of the e-sRGB space is about -0.5 to 1.67, but the hard copy is about 0.01 to 0.8. Therefore, in the three-dimensional lookup table for e-sRGB → CMYK conversion, a table is created so as to perform color gamut compression including the dynamic range. A method for creating a three-dimensional lookup table will be described below with reference to FIG.

  First, in step S21, the e-sRGB signal is converted into XYZ tristimulus values according to equations (3) to (5). Next, in step S22, the XYZ tristimulus values are converted into lightness J, saturation C (or M), and hue h in the CIECAM02 space. The specific calculation procedure of the XYZ⇒JCh conversion is described in Non-Patent Document 2 and the like, and thus the description thereof is omitted here.

  Next, the process proceeds to step S23, and gamut mapping processing for converting the JCh signal calculated in step S22 into a J'C'h 'signal within the color reproduction range that can be output by the output device is performed. As a procedure of gamut mapping processing, first, lightness compression is performed, and then saturation compression is performed so that it falls within the color reproduction range of the output device. Here, when the brightness J obtained in step S22 exceeds the reference white color (J = 100), the input brightness is compressed to the output brightness range by linear conversion of the brightness. In the saturation compression, for example, it is determined whether the output device is within or outside the reproduction range, and if it is outside, mapping is performed to the color having the smallest color difference among the color signals that can be reproduced by the output device. Further, the mapping method is not limited to the minimum color difference, and various methods such as storing and compressing gradation can be used.

  By the gamut mapping process in step S23, the signal is converted into a CIECAM02 signal within the color reproduction range of the output device, and then in step S24, CAM inverse conversion is performed. The conversion formula of the inverse conversion of CAM is described in Non-Patent Document 2 described above, and will be omitted. However, parameters suitable for the viewing conditions of the output device are used as appearance parameters used at the time of inverse transformation. For example, when converting to a printer signal, the following appearance parameters are used.

Adaptation luminance La = 64
Reference white point Xw, Yw, Zw: Tristimulus value of paper white Relative luminance of background area Yb: 20%
Ambient condition average
Finally, in step 25, XYZ is converted to CMY.

  In the above description, a printer is used as an example of the output device, but conversion to an sRGB signal for monitoring can be easily handled by replacing the color correction unit.

Example 2:
FIG. 12 shows a system configuration example when the present invention is realized by software. In this image processing system, a computer 300 such as a workstation, a printer 302, and a display 301 are connected. The workstation (computer 300) realizes the functions of the image processing method described above, and includes a display 301, a keyboard 309, a program reading device 310, an arithmetic processing device, and the like. In the arithmetic processing unit, a ROM 305 and a RAM 304 are connected to a CPU 303 capable of executing various commands via a bus. Further, a DISK 306 such as a hard disk, which is a large-capacity storage device, and a NIC 307 that communicates with devices on the network are connected to the bus.

  The program reading device 310 is a storage medium storing various program codes, that is, a floppy disk, a hard disk, an optical disk (CD-ROM, CD-R, CD-R / W, DVD-ROM, DVD-RAM, etc.), optical An apparatus for reading a program code stored in a magnetic disk, a memory card, or the like, such as a floppy disk drive, an optical disk drive, or a magneto-optical disk drive.

  The program code stored in the storage medium is read by the program reader and stored in the DISK 306 and the like, and the program code stored in the DISK 306 and the like is executed by the CPU 303, thereby realizing the functions of the image processing method described above. Will be able to. In addition, by executing the program code read by the computer, an OS (operating system) or a device driver running on the computer performs part or all of the actual processing based on the instruction of the program code. The case where the above-described function is achieved by the processing is also included.

1 shows a configuration of an image processing apparatus of the present invention. The example of a backlight scene and a brightness | luminance histogram is shown. 1 shows a configuration example of an image processing system. The relationship diagram of color space is shown. Indicates the range of each color space. The flowchart of a color distribution analysis method is shown. The flowchart of a range correction process is shown. An example of a range correction table is shown. The conversion characteristic of sc-RGB signal and e-sRGB signal is shown. An example of the background / subject separation result is shown. An example of range correction is shown. The flowchart of the preparation method of a color correction parameter is shown. An example of a system configuration when the present invention is implemented by software is shown.

Explanation of symbols

101 range compression unit 102 color correction unit 103 image analysis unit

Claims (10)

  An image processing method for converting an input color signal in a scene reference color space into a color signal for an output device, analyzing a color distribution of the input color signal, and determining a gradation compression condition based on the analysis result According to the gradation compression condition, the input color signal is mapped to an extended color space that is narrower than the scene reference color space and includes the color reproduction range of the output device, and the color signal mapped to the extended color space is output. An image processing method comprising color-converting to a color signal for a device.   The image processing method according to claim 1, wherein a color space that can be converted from a color signal expressed in a scene reference color space using a one-dimensional lookup table is used as the extended color space.   2. The image processing method according to claim 1, wherein, in the color distribution analysis, at least a luminance of a subject region and a luminance range of an input color signal are extracted.   As a result of the color distribution analysis, when the input color signal does not include a person region, the dynamic range of the input color signal is compressed by performing linear luminance conversion with reference to the luminance range of the input color signal. 2. The image processing method according to claim 1, wherein each component value of the compressed color signal is converted into a color signal in an extended color space by performing nonlinear conversion.   As a result of the color distribution analysis, if the input color signal includes a person area, whether the brightness of the person area is appropriate based on the ratio between the average brightness value of the person area and the maximum brightness value of the input color signal If it is determined as appropriate, the dynamic range of the input color signal is compressed by performing linear luminance conversion with reference to the luminance range of the input color signal, and the compressed color signal 2. The image processing method according to claim 1, wherein each of the component values is converted into a color signal in an extended color space by performing nonlinear conversion.   As a result of the color distribution analysis, if the input color signal includes a person area, whether the brightness of the person area is appropriate based on the ratio between the average brightness value of the person area and the maximum brightness value of the input color signal The subject region and the non-subject region are separated if the judgment is not appropriate, and a compression process with a different dynamic range is performed for each pixel with reference to the separation result. The image processing method as described.   The image processing method according to claim 6, wherein in the dynamic range compression processing, conversion characteristics are set using a luminance distribution characteristic in the vicinity of the target pixel and an attribute of the target pixel.   An image processing apparatus for converting an input color signal in a scene reference color space into a color signal for an output device, a means for analyzing a color distribution of the input color signal, and a gradation compression condition based on the analysis result And a means for mapping the input color signal to an extended color space that is narrower than a scene reference color space and includes a color reproduction range of an output device according to the gradation compression condition, and mapping to the extended color space An image processing apparatus comprising: a color conversion unit that converts the color signal into a color signal for an output device.   A program for causing a computer to implement the image processing method according to any one of claims 1 to 7.   A computer-readable recording medium recording a program for causing a computer to implement the image processing method according to claim 1.

Images