JP2009010636A - Adaption histogram equalization method, and adaption histogram equalization apparatus - Google Patents

Abstract

<P>PROBLEM TO BE SOLVED: To provide an adaption histogram equalization method, an adaption histogram equalization apparatus, and a computer program product for emphasizing contrast in an image, in particular, a technology for preventing occurrence of fading artifact and object magnification artifact caused by adaption histogram equalization. <P>SOLUTION: The adaption histogram equalization method for emphasizing contrast in a digital image includes a step for dividing the image into a region of pixels and deciding the structure for difference between local pixel values of a predetermined strength in the image, a step for generating a histogram of pixel values for each region based on the decided structure for the difference between local pixel values, and a step for mapping pixel values of each region based on the histogram corresponding to the region. <P>COPYRIGHT: (C)2009,JPO&INPIT

Description

  The present invention relates to an adaptive histogram equalization method, an adaptive histogram equalization apparatus, and a computer program product for enhancing image contrast. In particular, the present invention relates to a technique for preventing fading artifacts and object magnification artifacts from being generated by adaptive histogram equalization.

  Histogram equalization is an effective process for improving the contrast of an image. Histogram equalization, as a basic principle, equalizes the distribution of pixel intensity histograms in an image and improves the contrast of the image.

  The histogram equalization is usually performed as follows. First, a histogram is generated from the intensity value of each pixel of the image. After generating the histogram, the histogram is processed by leveling the histogram. Then, a cumulative function is derived from the histogram. This is usually a cumulative histogram. Then, a mapping function is generated from this cumulative function. Here, the accumulation function is usually normalized and an offset is applied. Based on the mapping function, the pixel intensity value is mapped to a new value. As a result of this processing, an image in which the histogram is more evenly distributed, that is, the contrast is enhanced is generated.

  This process may be applied to the entire image at once, or it may be divided into subsections called regions or blocks, and the process may be applied individually to these subsections. Subsections may overlap. A technique for applying histogram equalization to the whole is called global histogram equalization, and a technique for applying histogram equalization for each subsection is called local histogram equalization or adaptive histogram equalization. Local histogram equalization is adapted to local image features, and global histogram equalization may not work well for such local features. Local histogram equalization is usually much more computationally intensive than global histogram equalization, but has attracted attention in the field of image contrast enhancement due to its superior performance.

  Local histogram equalization techniques are described in, for example, “Digital Image Processing”, 3rd Edition, ISBN 0-471-37407-5, John Wiley & Sons, Inc., 2001 by William K. Pratt. Is described.

  Local histogram equalization can cause annoying artifacts such as:

  Blocking artifact: In the processed image, the block structure used for processing for each block is visually recognized, and in particular, the block boundary is conspicuous.

  Fading artifact: An artifact that fades some areas of the image after adaptive histogram equalization. Usually, the area in the vicinity of an object with high contrast is affected by this artifact. Human vision is more sensitive to skin color (human skin color) objects than general objects, and this artifact is particularly uncomfortable when it affects the skin color area.

  Boundary smearing artifact: An artifact where the boundary area of high contrast objects (eg, dark and light objects) becomes “smeared” or smeared and thicker than before processing. Boundary smear artifacts are also referred to as object magnification artifacts.

  Total image contrast loss: Local processing ignores the overall histogram information and may reduce the overall contrast. For example, local histogram equalization may make the background of the image as clear as the target object on the image. By emphasizing the background in this way, the human sense of distance may be confused.

  European patent application EP1515274 by the same applicant as this application discloses a technique for solving the problems of blocking artifacts and overall image contrast loss. Blocking artifacts and overall image contrast loss are suppressed by homogenizing the histograms of the blocks, ie, by weighting and summing the histograms of spatially adjacent blocks. Although the processing for overlapping blocks can also reduce blocking artifacts and overall image contrast loss, histogram equalization is more computationally efficient than the processing for overlapping blocks. A method of overlapping blocks and histogram equalization may be combined.

  A European patent application EP 1465536 by the same applicant as this application discloses a technique for solving fading artifacts. In the technique disclosed in EP 1465436, a histogram is generated from luminance components of an image signal such as a television signal. Then, the skin color region is specified using the color difference component of the image signal. Then, the detected skin color area pixels are used to change the histogram so that the histogram is more evenly distributed, thereby preventing the signal from changing at all or hardly in the human skin color area by histogram equalization. . Pixels in areas other than the skin color are used to generate a histogram similar to the standard histogram equalization.

  An object of the present invention is to provide an adaptive histogram equalization method and an adaptive histogram equalization apparatus that can reduce object enlargement artifacts and fading artifacts due to adaptive histogram equalization.

  In order to achieve this object, an adaptive histogram equalization method according to the present invention is an adaptive histogram equalization method for enhancing the contrast of a digital image. The adaptive histogram equalization method divides an image into pixel regions, Determining a pixel value difference structure, generating a pixel value histogram for each region based on the determined local pixel value difference, and each region based on a histogram corresponding to the region; Mapping the pixel value of the region.

  In order to achieve this object, an adaptive histogram equalization apparatus according to the present invention is an adaptive histogram equalization apparatus that enhances the contrast of a digital image, an area identifier that divides an image into pixel areas, A structure determiner that determines a structure of a difference between local pixel values of a predetermined intensity, a histogram generator that generates a histogram of pixel values based on the structure determined as a difference of local pixel values for each area, and an area And a mapper for mapping the pixel values of the respective regions based on the histogram corresponding to.

  Furthermore, in order to achieve this object, the software product according to the present invention includes program information that, when executed by a computer device, implements the adaptive histogram equalization method described above.

  The adaptive histogram equalization apparatus, adaptive histogram equalization method, and software product according to the present invention reduce fading artifacts and boundary smear artifacts. The present invention reduces fading artifacts in the flesh-colored area without the need to access color information, and also reduces fading artifacts in non-skin-colored areas.

  The adaptive histogram equalization method is preferably performed as follows.

  Preferably, for each region having a determined structure of local pixel value differences, a first side and a second side opposite the first side with respect to the determined structure are defined.

  For regions having a determined structure of local pixel value differences, a histogram is preferably generated from the pixel values of the pixels located on the first side of the determined structure.

  Preferably, for a region having a structure in which the difference between local pixel values is determined, the histogram is changed according to the number of pixels located on the second side of the structure, and the more pixels located on the second side, the more the histogram The degree of uniformity of the balance or distribution of the histogram generated by the step of generating.

  The pixels located on the second side are preferably considered in the histogram as if some of the pixels have a minimum value and other parts have a maximum value.

  The adaptive histogram equalization method preferably further comprises the step of leveling the histogram, or using the leveling function obtained from the histogram directly or indirectly in the step of mapping the pixel values.

  In regions that do not have a determined local pixel value difference, the histogram is preferably generated from the pixel values of all the pixels in the region.

  Preferably, for example, the structure of the difference between local pixel values having a predetermined intensity is determined using an edge detection method including a Sobel operator.

  Preferably, a single sparsely distributed edge point determined by the edge detection method is deleted and / or discontinuous edges are connected, for example by a morphological filter.

  The adaptive histogram equalization method determines a skin color pixel, changes the histogram according to the number of skin color pixels in an area having skin color pixels, and generates a histogram as the number of skin color pixels in the area increases. The degree of balance or distribution uniformity of the generated histogram is high.

  The histogram is preferably uniformized.

  The regions preferably overlap, and the values of the pixels located in the overlapping part of the region and the values of the pixels located in the non-overlapping part of the region are mapped only once.

  The pixel value is preferably a value related to the lightness and / or chromaticity of the pixel, including the luminance and / or color difference or chrominance difference value of the image format based on YUV, YIQ, YCbCr or other color model.

  The adaptive histogram equalization device is preferably realized as follows.

  Preferably, for each region having a determined structure of local pixel value differences, a first side and a second side opposite the first side with respect to the determined structure are defined.

  Preferably, for a region having a structure for which the difference between local pixel values is determined, the histogram generator generates a histogram from the pixel values of the pixels located on the first side of the determined structure.

  For regions having a structure with a local pixel value difference determined, the histogram generator preferably changes the histogram according to the number of pixels located on the second side of the structure, and the number of pixels located on the second side The greater the number, the higher the degree of uniformity or balance of the histogram generated by the histogram generator.

  The histogram generator preferably regards the pixels located on the second side in the histogram as if some of the pixels have a minimum value and other parts have a maximum value.

  The adaptive histogram equalizer preferably further comprises a filter for leveling the histogram, or the mapper maps the pixel values directly or indirectly using the leveling function obtained from the histogram.

  The histogram generator preferably generates a histogram from the pixel values of all the pixels in the region in the region that does not have the structure in which the difference between the local pixel values is determined.

  The structure determiner preferably determines a difference structure of local pixel values having a predetermined intensity using an edge detection method including a Sobel operator.

  The structure determiner preferably removes a single sparsely distributed edge point determined by the edge detection method and / or discontinuous edges are connected by a morphological filter.

  The adaptive histogram equalization processor preferably further comprises a skin color detector for determining skin color pixels, and in the region having skin color pixels, the histogram generator changes the histogram according to the number of skin color pixels, and the skin color of the region The greater the number of pixels, the higher the degree of uniformity in the balance or distribution of the histogram generated by the histogram generator.

  The histogram is preferably uniformized.

  The regions preferably overlap, and the mapper maps the values of the pixels located in the overlapping part of the region and the values of the pixels located in the non-overlapping part of the region only once.

  The pixel value is preferably a value related to the lightness and / or chromaticity of the pixel, including the luminance and / or color difference or chrominance difference value of the image format based on YUV, YIQ, YCbCr or other color model.

  If the input image is a digital image, the present invention can be applied most easily. If the input image is not a digital image, the image format is converted from an analog format to a digital format before the image contrast enhancement processing. The input image may be a color image or a black and white image. The input image may be a single image or an image sequence. In other words, the present invention can be applied to still images such as photographs and image sequences such as video images. For example, if the input image format includes data relating to the lightness of the pixels of the image, for example, grayscale values of a grayscale scanned image, or includes, for example, data indicating the luminance of the video signal of a standard television signal In this case, the present invention can be applied to the input image. Such lightness data is obtained from, for example, the red / green / blue (RGB) color model before the lightness data is used, YUV, YIQ used in PAL and NTSC, respectively, and the “television standard”. It can be generated from data in the input format at any time by conversion to a luminance color difference color model such as a YCbCr color model. Further, a structure of a difference between local pixel values having a predetermined intensity is determined based on the pixel value, and a different type of value from the pixel value for which the histogram is generated can be mapped. For example, if a histogram is generated based on the local green pixel value difference structure of a predetermined intensity, the pixel value related to lightness is mapped based on the histogram generated from the pixel value related to green May be. Since the green component usually has a good S / N ratio, processing based on the green component of the color signal is advantageous. The present invention can also be applied to enhance the contrast of data not related to lightness, and this data can be color related data or any other data.

  The present invention is realized as an adaptive histogram equalization method, an adaptive histogram equalization apparatus, or a computer program product. In the following, for the sake of brevity, the adaptive histogram equalization method according to the present invention will be mainly described. However, the disclosure according to the present specification describes the corresponding adaptive histogram equalization apparatus and the adaptive histogram equalization method. The same applies to the computer program product to be realized.

  FIG. 1 shows a flowchart of an adaptive histogram equalization method according to the present invention. In step S2, the image is divided into pixel regions. In step S4, the structure of the local pixel value difference of a predetermined intensity of the image is determined. In step S8, a histogram of pixel values is generated based on the determined structure of local pixel value differences in each region. In step S12, the pixel value of each region is mapped based on the histogram corresponding to the region. As an option, a skin color pixel may be determined in step S6. The generation of the histogram in step S4 may be performed based on the determined skin color pixels. In step S10, the histogram of each region may be leveled.

  These steps may be executed individually for each region. This means that, for example, a histogram of another region is generated in a state where pixel values of one region are already mapped, and the other regions may not be defined.

  FIG. 2 shows a block diagram of the corresponding adaptive histogram equalization processor 1. The adaptive histogram equalization processor 1 includes an area identifier 2 that divides an image into areas, a structure determiner 4 that determines a structure of a difference between local pixel values of a predetermined intensity of the image, and a local pixel value for each area. A histogram generator 8 that generates a histogram of pixel values based on the difference determined structure, and a mapper 12 that maps the pixel values of each region based on the histogram corresponding to the region. Further, the adaptive histogram equalization processor 1 may include a skin color detector 6 that detects skin color pixels. In this case, the histogram generator 8 generates a histogram based on the determined structure of local pixel value differences and based on the determined skin color pixels. Furthermore, the adaptive histogram equalization processor 1 may include a filter 10 for leveling the histogram.

  The area specifying unit 2, the structure determining unit 4, the skin color detector 6, the histogram generator 8, the filtering unit 10, and the mapping unit 12 of the adaptive histogram equalization processor 1 are realized by any arithmetic device or any device that performs information processing. Also good. For example, a suitably programmed personal computer, a digital signal processor based computing device suitable for image processing, or any other microprocessor based information processing device may be used. The adaptive histogram equalization processor 1, the region identifier 2, the structure determiner 4, the skin color detector 6, the histogram generator 8, the filtering device 10, and the mapping device 12 may be realized by the same hardware or different hardware. It may be realized. Hereinafter, an embodiment of the adaptive histogram equalization method according to the present invention will be described with reference to FIGS.

  In a practical implementation of the adaptive histogram equalization processor 1, further elements are required including, for example, a housing, a power supply, a data interface, a data connection terminal, a buffer and the like. These elements necessary for a practical implementation will be apparent to those skilled in the art and will not be discussed here.

  For the sake of clarity, it is assumed that the input image processed in the embodiment of the present invention is a digital format image including luminance data. The pixel region is called a block and has a rectangular shape. A shape other than a rectangle may be defined. The shape and position of the region are independent of the image content, and in particular, independent of the local pixel value difference structure. Details regarding the process of dividing the image into pixel regions will be described later. Processing for each block is performed from left to right and from top to bottom. This order corresponds to the temporal order in which standard television and video signals carry image information. Processing the signals in the natural order of the signals is preferable because the processing is faster and the required memory capacity is reduced. Many other block processes are possible, for example, blocks may be processed at least partially in parallel.

  FIG. 3 shows a flowchart of the central part of the embodiment of the present invention. In the configuration shown in FIG. 3, the process from the determination of the difference structure of local pixel values to the process of generating a histogram including histogram leveling is performed.

  In step S20, a block having a local luminance difference structure with a predetermined intensity is determined by an edge detection method. The sensitivity of edge detection can be adjusted by an edge threshold value parameter. The predetermined intensity usually defines a lower limit of the intensity of the local pixel value difference, and the intensity may be measured by any suitable technique. As described above, an object with high contrast usually causes fading artifacts and object magnification artifacts. In a typical embodiment of the present invention, the determined local pixel value difference appears larger or stronger than the local pixel value difference corresponding to the gradual change. The structure determined in this sense is a structure of strong local pixel value differences. In the present invention, a low pixel value difference may be used. More accurate values of the intensity of the local pixel value difference are rarely available. For example, image characteristics, such as the amount of image noise or the type of image motif, affect the detection of usable values. Furthermore, in some applications, such as high quality video post production where the edge parameters are adjusted by a human operator, this value may depend on the operator's personal preference. Also, in the processing of digital photographs in computer-based applications that are currently popular at the amateur level, the user can also directly change the intensity value. As another specific example, a camcorder user may select a predetermined range of preset values, each optimized for a typical shooting situation, in particular the intensity of the pixel value difference to be determined. May be a value for setting. As another possibility, for example, automatic setting may be performed based on shooting conditions such as lighting conditions, or other information previously acquired with respect to an image, for example, information acquired by a previous image processing operation. is there. Also, different values may be set for different portions of the image, for example, different regions.

  For the sake of explanation, the acquired information can also be displayed as an image. The image of FIG. 6 is an image obtained by processing the entire image of FIG. 5 using a known Sobel operator. Here, the structure for which the difference between the local pixel values is determined is displayed in white on a black background. The structure in which the difference between local pixel values is determined can also be called an edge, an edge line, an edge point, a boundary, a boundary line, boundary information, or the like. As can be seen from these figures, it can be seen that the structure of the difference between the local pixel values matches the boundary line of the object having a high contrast. The structure of the difference between local pixel values is usually not limited to any shape.

  In step S22 following step S20, it is determined whether or not an edge exists in the block. If not, the process proceeds to step S30. On the other hand, if applicable, the process proceeds to step S24.

  In step S24, in the histogram, the luminance of the pixel behind the edge is alternately regarded as a minimum value and a maximum value, for example, a value 16 and a value 235 in 8-bit quantization of a standard television signal. The minimum and maximum values may differ to some extent from the minimum and maximum values actually possible in a given image format, as long as each is a relatively small and large value within the possible dynamic range.

  Whether a pixel is behind or in front of an edge may be determined based on the processing order, but usually the role of the pixel before the edge and the role of the pixel behind the edge are interchangeable. is there. Here, for the convenience of explanation, the expression pixel before or after the edge is used.

  In step S26, it is determined which type of block end has been reached. If the block end in the horizontal x direction has been reached, the process proceeds to S28, the next line is specified, and the process proceeds to step S22. On the other hand, if the block end in the vertical y direction has been reached, the process proceeds to step S30.

  When the process proceeds from step S28 to step S22, it is determined in step S22 whether or not a further edge exists. If so, the process proceeds to step S24. If not, the process proceeds to step S30.

  In step S30, a histogram is generated by directly taking into account the pixel value of the pixel in front of the edge. The histogram is completed by the process in step S30, and the process proceeds to step S32.

  As described above, in the picture area where the local contrast is not strong, the histogram is generated as usual. On the other hand, in a picture region having a high local contrast, only a pixel before the detected object boundary is directly taken into consideration to generate a histogram. The appearance of other pixels is not directly taken into account, but by alternately counting the minimum and maximum values of the histogram until it reaches the block edge, the histogram is made more uniform, or the histogram is made more uniform Used as an index to change to act as a histogram. The alternating count continues until both the horizontal (x) block edge and the vertical (y) block edge are reached. The histogram takes only the minimum value and the maximum value from the edge start position to the block end.

  In step S32, the histogram is processed. This processing includes, for example, histogram leveling including filtering by a low-pass filter. The histogram obtained in this way may be weighted and added to the histogram previously obtained for adjacent blocks. This is one simple technique for histogram equalization. In many cases, histogram equalization reduces artifacts caused by single edge points, sparsely distributed edge points, or discontinuous edges. As an alternative to or in addition to histogram equalization, as part of step S20, single edge points or sparsely distributed edge points are removed and discontinuous using, for example, a known morphological filter Various edges may be connected.

  After step S32, a mapping function is derived from the cumulative function of the leveled histogram, and the luminance of the block is mapped based on this mapping function. In this embodiment, not all the pixels of the block are mapped. This point will be described in detail later. Once all blocks have been processed and the corresponding pixels mapped, histogram equalization for the input image is complete.

  Alternatively, by alternately counting the minimum and maximum values, the balance of the histogram is balanced, especially after leveling the histogram, the histogram becomes more uniform. When the histogram is more uniformly distributed, the change in the pixel value due to the mapping using the mapping function derived from the histogram becomes smaller. In this way, the amount of histogram equalization is reduced for blocks containing edge points. In other words, the amount of histogram equalization is reduced in an area including an object with high contrast. As a result, fading artifacts and boundary smear artifacts are suppressed. In areas without edges, the histogram is generated as usual. Therefore, histogram equalization is performed with the maximum amount in the region excluding objects with high contrast. Instead of leveling the histogram, the cumulative function derived from the histogram may be leveled, and this effect is the same as when the amount of histogram equalization is reduced.

  Pixels before the boundary are taken into account directly in the generation of the histogram, and pixels after the boundary are changed to make the histogram more uniform or to behave as a more uniform histogram. Is used as an index. In this embodiment, this process is performed by alternately counting the minimum and maximum luminance values until the end of the block is reached. Thereby, the balance of the histogram is balanced, and after the histogram is leveled, the histogram becomes more uniform as compared with a case where the histogram is completely generated from actual pixel values. Therefore, as a result of the equilibrium, the histogram behaves like a more uniformly distributed histogram. Histograms can be generated in other ways, with similar effects. Also, uniformity can be achieved more directly. For example, pixels that do not take actual values into account may be used to fill histogram gaps constructed from other pixels that take into account actual values, thereby directly increasing histogram uniformity.

  If the pixels before the detected boundary are the majority of the block, the histogram change has little or no effect on the final result. In such cases, empirically, fading artifacts and object magnification artifacts are often very weak, i.e., not visible. On the other hand, when the ratio of the pixels before the boundary to the block decreases, fading artifacts and object enlargement artifacts become conspicuous, but the histogram change is also strongly performed, thereby suppressing the artifacts. In this way, artifacts are reduced based on the actual required amount.

  Although the present invention has been described based on specific embodiments, it is clear that the roles of the two sides may be reversed, i.e., to make the histogram more uniform or to make the histogram more uniform. The pixels before the boundary line may be used as an index to change to behave as a histogram, and the pixels after the boundary line may be directly taken into account to generate the histogram.

  Hereinafter, the block structure used by the embodiment of the present invention will be described with reference to FIG. The present invention can be used in combination with conventional techniques for overlapping blocks. Block overlap is known as a technique for reducing blocking artifacts. However, block duplication can intensify fading artifacts and object magnification artifacts. FIG. 4 shows three rectangular blocks densely overlapping in the horizontal x direction. FIG. 4 also shows three individually separated blocks (lower three blocks) for visual clarity. It is necessary to map only the pixels in the region filled with the pattern shown in FIG. 4 based on the mapping function derived from the entire block so that the mapping does not overlap. The block structure used in this embodiment also overlaps in the vertical direction. However, for the sake of clarity, FIG. 4 does not show vertical overlap of blocks. Therefore, the area of the region mapped by the mapping function needs to be further reduced so that the mapping does not overlap. That is, mapping using a mapping function derived from the entire block is performed only in regions that do not overlap both horizontally and vertically. Specifically, in this embodiment, only the area of the upper left corner of each block is mapped. That is, the histogram is generated from the entire block, but only a very small portion of each block's pixels are mapped. Since only the pixels in the upper left corner are mapped, the histogram change usually does not affect the pixels behind the detected boundary.

  FIG. 7 shows two difference images between the processed image and the original image (here, a predetermined offset value is set so that the difference is visually clear in the printed picture). Are added and amplified by a predetermined coefficient). In the right picture, the original image has been processed using an embodiment of the histogram equalization method according to the present invention. In the left picture, the original image has been processed using conventional histogram equalization. In the difference image on the left (based on the prior art), the brightness changes on the boundary line that separates the background from the shoulder and the arm, and a noticeable object enlargement artifact occurs. In the border region, the pullover is darker and the background is lighter by processing the image using conventional techniques. Further, at the boundary line between the pullover and the neck, a change in brightness corresponding to a fading artifact is also generated. That is, the skin color is lighter near the pullover. On the other hand, as is apparent from the picture on the right side, these artifacts are greatly reduced by performing processing based on the histogram equalization method according to the present invention.

  Although the preferred embodiment and modifications of the invention have been disclosed and described herein, various changes can be made in the arrangement, operation and form disclosed herein without departing from the spirit and scope of the invention. It will be apparent to those skilled in the art that this is possible. In particular, each feature of the present invention can be combined in any configuration, except where apparently meaningless to those skilled in the art, including features disclosed in combination with other features of the present invention.

It is a flowchart of the adaptive histogram equalization method including arbitrary steps. FIG. 3 is a block diagram of an adaptive histogram equalization processor that includes arbitrary elements. It is a flowchart explaining 1 side of embodiment of the adaptive histogram equalization method. It is a figure explaining the duplication of the horizontal direction of the block structure used in embodiment of this invention. It is a figure which shows the original image. It is a figure which shows the structure of the difference of the local pixel value of an original image. It is a figure which shows two difference images between the processed image and the original image.

Claims (27)

In an adaptive histogram equalization method that enhances the contrast of a digital image,
Dividing the image into regions of pixels and determining in the image the structure of the difference between local pixel values of a predetermined intensity;
Generating a histogram of pixel values based on the determined structure of local pixel value differences for each region;
Mapping the pixel value of each region based on the histogram corresponding to the region.   A first side and a second side opposite to the first side with respect to the determined structure are defined for each region having the determined structure of the local pixel value difference. The adaptive histogram equalization method according to claim 1.   The region having a structure for which the difference between the local pixel values is determined is characterized in that the histogram is generated from pixel values of pixels located on the first side of the determined structure. Item 3. The adaptive histogram equalization method according to item 2.   For a region having a structure in which the difference between the local pixel values is determined, the histogram is changed according to the number of pixels located on the second side of the structure, and the more pixels located on the second side, 4. The adaptive histogram equalization method according to claim 2, wherein the degree of uniformity of the balance or distribution of the histogram generated by the step of generating the histogram is high.   5. The pixel located on the second side is regarded in the histogram as if a part of the pixel has a minimum value and another part has a maximum value. Adaptive histogram equalization method.   The leveling of the histogram is further included, or the leveling function obtained from the histogram is directly or indirectly used in the step of mapping the pixel values. The adaptive histogram equalization method according to claim 1.   7. The histogram according to claim 1, wherein the histogram is generated from pixel values of all the pixels in the region in a region that does not have a structure in which a difference between local pixel values is determined. Adaptive histogram equalization method.   The adaptive histogram equalization method according to claim 1, wherein a structure of a difference between the local pixel values having a predetermined intensity is determined using an edge detection method including a Sobel operator.   9. The adaptive histogram according to claim 8, wherein single sparsely distributed edge points determined by the edge detection method are deleted and / or discontinuous edges are connected by a morphological filter. Equalization method.   Histogram generated by the step of determining a skin color pixel and changing the histogram according to the number of skin color pixels in the region having the skin color pixel, and generating the histogram as the number of skin color pixels in the region increases. 10. The adaptive histogram equalization method according to claim 1, wherein the degree of uniformity of the balance or distribution is high.   11. The adaptive histogram equalization method according to claim 1, wherein the histogram is uniformized.   The region is overlapped, and a value of a pixel located in an overlapping portion of the region and a value of a pixel located in a non-overlapping portion of the region are mapped only once. The adaptive histogram equalization method according to claim 1.   The pixel value is a value related to brightness and / or chromaticity of a pixel including luminance and / or color difference or chrominance difference value of an image format based on YUV, YIQ, YCbCr or other color models. The adaptive histogram equalization method according to claim 1. In an adaptive histogram equalizer (1) that enhances the contrast of a digital image,
An area identifier (2) for dividing an image into pixel areas;
A structure determiner (4) for determining a structure of a difference between local pixel values of a predetermined intensity of the image;
For each region, a histogram generator (8) that generates a histogram of pixel values based on the determined structure of local pixel value differences;
An adaptive histogram equalizer (1) comprising a mapper (12) for mapping pixel values of each region based on a histogram corresponding to the region.   A first side and a second side opposite to the first side with respect to the determined structure are defined for each region having the determined structure of the local pixel value difference. 15. The adaptive histogram equalizer (1) according to claim 14.   The histogram generator (8) generates a histogram from the pixel values of the pixels located on the first side of the determined structure with respect to the region having the determined structure of the local pixel value difference. 16. The adaptive histogram equalizer (1) according to claim 15, characterized in that   For the region having the structure for which the difference between the local pixel values is determined, the histogram generator (8) changes the histogram according to the number of pixels located on the second side of the structure, and the second side 17. The adaptive histogram equalization apparatus (1) according to claim 15 or 16, characterized in that the greater the number of pixels located at, the higher the degree of uniformity of the balance or distribution of the histogram generated by the histogram generator (8). ).   The histogram generator (8) regards the pixel located on the second side in the histogram as if part of the pixel has a minimum value and the other part has a maximum value. The adaptive histogram equalization apparatus (1) according to claim 17, characterized in that:   It further comprises a filter (10) for leveling the histogram, or the mapper (12) maps the pixel values directly or indirectly using the leveling function obtained from the histogram. 19. The adaptive histogram equalizer (1) according to any one of claims 14 to 18.   15. The histogram generator (8) generates a histogram from pixel values of all pixels in the region in a region that does not have a structure in which a difference between local pixel values is determined. 19. The adaptive histogram equalization apparatus (1) according to any one of 19 items.   21. The structure determination unit according to claim 14, wherein the structure determination unit determines a structure of a difference between the local pixel values having a predetermined intensity using an edge detection method including a Sobel operator. Adaptive histogram equalizer (1).   The structure determiner removes a single sparsely distributed edge point determined by the edge detection method and / or discontinuous edges are connected by a morphological filter. 21. The adaptive histogram equalizer (1) according to item 21.   A skin color detector for determining skin color pixels is further provided. In the region having the skin color pixels, the histogram generator (8) changes the histogram according to the number of skin color pixels, and the number of skin color pixels in the region 23. The adaptive histogram equalization processor (1) according to any one of claims 14 to 22, characterized in that the greater the number of the histograms, the higher the degree of uniformity of the balance or distribution of the histograms generated by the histogram generator (8). .   24. The adaptive histogram equalization processor (1) according to any one of claims 14 to 23, characterized in that the histogram is uniformized.   The region overlaps, and the mapper maps the value of a pixel located in an overlapping portion of the region and the value of a pixel located in a non-overlapping portion of the region only once. 25. The adaptive histogram equalization processor (1) according to any one of items 24 to 24.   The pixel value is a value related to brightness and / or chromaticity of a pixel including luminance and / or color difference or chrominance difference value of an image format based on YUV, YIQ, YCbCr or other color models. The adaptive histogram equalization processor (1) according to any one of claims 14 to 25.   A software product comprising program information that, when executed by a computer device, implements the adaptive histogram equalization method according to any one of claims 1 to 13.

Images