JP2007334675A - Device, method and program for processing image, and recording medium - Google Patents

Abstract

<P>PROBLEM TO BE SOLVED: To solve the problem that an area to be extracted cannot be completely detected by the detection using one representative chromaticity. <P>SOLUTION: An image processor for detecting a specific characteristic area from the image data comprises a first detection means for detecting a candidate of the characteristic area based on the preset chromaticity range, a re-setting means for re-setting the chromaticity range based on chromaticity information of the candidate of the characteristic area detected by the first detection means, a second detection means for further detecting a candidate of the characteristic area based on the chromaticity range reset by the re-setting means, and a characteristic area determination means for determining whether the candidate of the characteristic area detected by the first or second detection means is the specific characteristic area. <P>COPYRIGHT: (C)2008,JPO&INPIT

Description

  The present invention relates to an image processing method and apparatus, and more particularly to a method of recognizing a feature region in image data based on chromaticity, an application method thereof, and an apparatus thereof.

  In recent years, input devices such as digital cameras have become widespread as image input devices. An image captured by such an input device may include a captured image that does not necessarily have a good balance due to color covering, inappropriate exposure, inappropriate contrast, and the like. Therefore, when viewing a photographed image or printing it for viewing, it may not be appropriate to use the photographed image data as it is.

  Therefore, an application that can define a good image and automatically perform white balance correction and exposure / contrast correction in accordance with the definition has become common. However, there may be a captured image that cannot always be corrected appropriately even under such an environment.

  In some cases, there are important areas of interest to the viewer in the image, apart from the balance of the entire image, and the image remains unsatisfied if the area is not well-drawn. turn into. Such an image cannot be properly corrected by the above method. In addition, as an important area | region, a human face, skin, etc. are mentioned, for example. In order to perform image correction in consideration of such an important area, the following method has been proposed.

  In the technique disclosed in Patent Document 1, in Jpeg image data, a detection target region in an image is determined and detected based on DCT characteristics, chromaticity, and the like of the data compression unit. For the 8 * 8 pixel data of the compression unit, which is the determination unit, the chromaticity is determined using the chromaticity of a specific pixel as a representative value, and this representative value is used as the chromaticity of the entire 8 * 8 pixel. Yes. This determination is made for a group of 8 * 8 pixels in the entire image.

  In the technique disclosed in Patent Document 2, the area of a skin color portion in one frame is detected from a color difference signal of a video signal, the area is set as a human skin area, and used as a target setting value for luminance correction, and gamma correction is performed. A circuit control signal is created. The determination of chromaticity is performed by detecting an area where the chromaticity signal of the video signal falls within the set chromaticity range. Therefore, the unit of chromaticity determination is for each pixel.

JP 2004-38480 A Japanese Patent No. 3203946

  As described above, when printing an image (such as a Jpeg image) taken with an input device such as a digital camera, it is necessary to be able to print a noticed image of a person or the like better like a silver halide photograph print. It is desirable that correction can be performed accordingly. In order to realize this, a method for finding a noticed image from a photographed image is required. On the other hand, there is a need for a method in which the detection process can be performed as lightly as possible so that it can be used even in a device having a low data processing capability such as direct printing that directly prints from a digital camera to a printer.

  In the conventional method of detecting the presence or absence of an image of interest such as a person in the shooting data, and further detecting the skin area of the person, a determination based on color components such as chromaticity is added to the area detection process. There is. When this chromaticity determination is performed, the chromaticity defined as the detection reference is normally defined by the chromaticity distribution of human skin.

  However, it is known that there is a difference in the chromaticity of the subject itself in the shooting data to be detected, and that this chromaticity is affected by the light source state at the time of shooting. In this case, when the detection process by the chromaticity representative of the uniform is performed, there is a problem that the extraction target region may not be completely detected from the chromaticity determination result. The present invention aims to solve the above problems.

  In order to achieve the above object, an image processing apparatus according to a first aspect of the present invention detects a feature region candidate based on a preset chromaticity range in an image processing device that detects a specific feature region from image data. A first detecting unit; a resetting unit that resets the chromaticity range based on chromaticity information of candidate feature regions detected by the first detecting unit; and a color reset by the resetting unit Second detection means for further detecting the feature area candidates based on the degree range, and whether the feature area candidates detected by the first detection means or the second detection means are the specific feature areas And a characteristic region determining means for determining.

  According to a second aspect of the present invention, there is provided an image processing apparatus for detecting a feature region candidate based on a preset chromaticity range in an image processing device for detecting a specific feature region from image data. A reset determination unit that determines whether or not the chromaticity range needs to be reset based on chromaticity information of the feature region candidates detected by the first detection unit; and the chromaticity by the reset determination unit A resetting unit for resetting the chromaticity range based on chromaticity information of a candidate feature region detected by the first detection unit when it is determined that the range needs to be reset; Second detection means for further detecting feature area candidates based on the chromaticity range reset by the means, and the feature area candidates detected by the first and second detection means include the specific feature area Feature region determination to determine whether Characterized in that it comprises a stage.

  According to a third aspect of the present invention, in the image processing apparatus according to the first or second aspect, the detection processing in the first and second detection means is performed on the image data in units of a block having a specific size. And

  According to a fourth aspect of the present invention, in the image processing apparatus according to the third aspect of the present invention, the feature area candidates include a group of consecutive blocks detected as candidates.

  According to a fifth aspect, in the image processing apparatuses according to the first to fourth aspects, the resetting unit resets the chromaticity range for each feature region candidate.

  According to a sixth aspect of the present invention, in the image processing device according to the first to fifth aspects, the feature region determination means makes the determination based on a spatial frequency characteristic of the feature region candidate. .

  According to a seventh aspect, in the image processing apparatuses according to the first to sixth aspects, the characteristic area includes at least a human skin area.

  The eighth invention is the image processing device according to any one of the first to seventh inventions, further comprising an image correcting means and a correction information creating means, wherein the correction information creating means creates correction information for the feature region. The image correction means performs image correction processing using the generated correction information for the feature region, and performs image correction processing defined in advance for all the regions for the other regions.

  According to a ninth aspect of the present invention, in the image processing method for detecting a specific feature region from image data, a first detection step of detecting the feature region candidate based on a preset chromaticity range; A resetting step for resetting the chromaticity range based on chromaticity information of candidate feature regions detected in one detection step; and a feature region based on the chromaticity range reset in the resetting step. A second detection step of further detecting a candidate; and a feature region determination step of determining whether the feature region candidate detected in the first detection step or the second detection step is the specific feature region. It is characterized by that.

  According to a tenth aspect of the present invention, in the image processing method for detecting a specific feature region from image data, a first detection step of detecting the feature region candidate based on a preset chromaticity range; A reset determination step for determining whether or not the chromaticity range needs to be reset based on chromaticity information of the feature region candidates detected in one detection step; and the resetting of the chromaticity range in the reset determination step A reset step for resetting the chromaticity range based on the chromaticity information of the feature region candidates detected in the first detection step, and resetting in the reset step. A second detection step of further detecting feature region candidates based on the determined chromaticity range, and the feature region candidates detected in the first and second detection steps are the specific feature Characterized in that it comprises a characteristic region determining step of determining whether a band.

  An eleventh aspect of the invention is a program for causing a computer to realize the functions of the units included in the image processing apparatus according to any one of claims 1 to 8.

  The twelfth invention is a computer-readable recording medium in which the program according to the eleventh invention is recorded.

  According to the present invention, it is possible to accurately extract and detect feature regions individually as necessary. In addition, by providing a mechanism that further reflects the extraction results of feature regions in the correction of individual images, it is possible to obtain image data that cannot be corrected well by conventional image correction based on the general definition of good images. In contrast, it is possible to achieve good image correction individually.

Hereinafter, an embodiment of the present invention will be described with reference to the drawings.
FIG. 11 is a schematic block diagram showing a hardware configuration of an image processing apparatus according to an embodiment of the present invention.

  As shown in FIG. 11, the image processing apparatus of the present embodiment is achieved by a so-called personal computer configuration. Specifically, the configuration includes an input / output port 51, a mouse 52, a keyboard 53, a monitor 54, a control unit 55, a memory card drive 56, a hard disk 57, and a bus 58 for connecting these units.

  The input / output port 51 is a port for inputting / outputting data to / from the outside. The mouse 52 is a kind of pointing device and is an input device. The keyboard 53 is an input device for inputting numerical values and characters. The monitor 54 is an output device for performing various displays. For example, a liquid crystal monitor, a CRT monitor, an EL display, a plasma display, and a television receiver can be used. As the monitor 54, a so-called touch panel having the functions of the mouse 52, the keyboard 53, and the monitor 54 may be used.

  The control unit 55 is a processing device composed of a CPU, and controls the operation of the entire device. Based on input from the mouse 52 or the keyboard 53, the image data read by the input / output port 51 or the memory card drive 56 is read out. Various processes are also performed. The memory card drive 56 is an input / output device for reading image data recorded in the memory card and writing data. The hard disk 57 is a non-volatile storage device that stores various data and programs.

  In this embodiment, a personal computer is described as an example of the image processing apparatus, but any apparatus that can process an image may be used. For example, a similar embodiment is conceivable in a device that can handle images, such as a non-PC embedded device such as a multifunction printer or a photo direct printer that handles images, or a mobile phone.

  Next, the overall flow of processing performed in the above configuration is shown in FIG. 1, and the operation of the image processing apparatus of the present embodiment will be described.

  First, in step S101, image data read by the input / output port 51 or the memory card drive 56 and image data stored in the hard disk 57 are read based on an input operation from the mouse 52 or the keyboard 53.

  In step S102, the read image file is expanded. During the series of operations, the settings for the feature region detection in the image are confirmed for the acquired image, and if the set feature region definition is a human skin region in the image, the skin of the person in the image Processing to detect an area is performed. Note that the processing for detecting a face region made up of a person's skin region is disclosed in Japanese Patent Application Laid-Open Nos. 2004-038480, 2004-038481, and 2004-038482 by the inventors of the present application. In addition, a process for detecting a human face area in an image including a moving image in an entrance / exit management system or a monitoring system is known.

  In this step, at least the chromaticity data is used to determine the characteristics of the human skin area or the face area, the matched area is extracted, the information is held, and the feature amount of the extracted area is calculated and held. Here, at least the luminance average value and the size information including the number of pixels are included.

  Here, the case where the human skin region in the image is used as the feature region will be described in detail based on the detection processing flowchart of FIG.

  In step S1001, the human skin region candidate group detection process is performed on the image data input in step S101 by using a combination of chromaticity and arrangement in units of 8 * 8 pixels, which are JPEG image compression units. The detection of candidate groups in step S1001 is the central content of the present embodiment, and details will be described later.

  In step S1002, it is determined whether a candidate group has been detected in step S1001. If the detection fails, the process proceeds to step S1006 to set that the detection of the human skin area has failed [HM = “0000”] (where HM is an identifier indicating the success / failure of the human skin area detection) Is). If the detection is successful, the process proceeds to step S1003.

  On the other hand, if it is determined in step S1002 that the candidate group has been successfully detected, the process proceeds to step S1003, and the human skin area candidate group that is a desired feature area is set and stored in advance. Compared with the range of spatial frequency characteristics.

  In step S1004, it is determined whether the detected group of human skin area candidates is suitable as a human skin area that is a desired feature area. As a result of the determination, if it is determined that the human skin area is not a desired feature area, the process proceeds to step S1006. On the other hand, if it is determined to be a human skin area, the process proceeds to step S1005.

  In step S1006, the fact that the detection of the human skin area has failed is set [HM = “0000”]. In step S1005, the fact that the human skin area has been successfully detected is set [HM = “1111”]. Then, the feature amount data (described above) of the detected feature region is stored.

  Note that the present invention is not limited to the above-described feature region detection means / method, and can be replaced with a detection method based on other human face detection or personal face approval.

  Returning to FIG. 1, when it is determined in step S103 that the detection result “HM” in step S102 is “1111”, it is necessary to create a correction table using the feature region extracted from the image. Judge that there is. Then, a correction table is created based on the feature amount of the feature region.

  On the other hand, if it is determined in step S103 that the detection result “HM” in step S102 is “0000”, the process proceeds to step S104.

  In step S104, the presence / absence of a correction table created based on the feature amount of the feature area in step S103 is determined, and if it does not exist, image correction based on a definition common to all areas that does not reflect the detection result of the feature area Execute. On the other hand, when there is a correction table, the feature region portion is shared so that the created correction table can be used for image correction based on the common definition so as to reflect the detection result of the feature region. Image correction is executed after replacement with the definition correction table. Note that the image correction processing itself can use the same technique as in the prior art.

Next, the method for detecting candidate groups in S1001 in the extraction processing flowchart of FIG. 10 will be described in detail.
In this embodiment, 8 * 8 pixels, which are image compression units of a Jpeg image, are set as a minimum unit group for chromaticity determination. A human skin chromaticity determination method that can be most utilized as chromaticity determination used in the present embodiment will be described.

There are a plurality of methods for detecting the human skin chromaticity. As known,
1) A color in which the ratio of B (blue) / G (green) falls within the range of 0.7 to 0.8 and the ratio of R (red) / G (green) falls within the range of 1.4 to 1.8. Have a degree.
2) As shown in the conceptual diagram of FIG. 12, the skin color can be represented by a probability ellipse, and falls within this probability ellipse. The following formulas (1) to (3) are well known as formulas for obtaining the established ellipse.

here,

Are the average values of the skin color chromaticities r and g, and σ _r ² and σ _g ² are the corresponding variances. Also, η is a coefficient representing probability, and η = 2.0, η = 1.5, and η = 1.0 correspond to probability ellipses of 95%, 84%, and 68%, respectively.

  In the present embodiment, the chromaticity distribution range represented by the following equation (4) taking into account the simplicity of processing is set as the skin color chromaticity range, and this is used as the chromaticity distribution for detecting the human skin region. The judgment range. This range is shown in FIG.

  Next, in order to determine the chromaticity of the group of 8 * 8 pixels, which is the image compression unit of the Jpeg image set as the minimum unit group for chromaticity determination, it is compared with the chromaticity range of equation (4). Consider how to set the chromaticity representative value for a group of 8 * 8 pixels.

  An example is shown in FIG. FIG. 7 illustrates chromaticity detection points used in this embodiment. According to this, it is confirmed whether all the chromaticities at the four corners of the 8 * 8 pixel unit block are within the chromaticity range of the equation (4). It is determined that there is.

  In FIG. 7, the second chromaticity from the left in the upper stage and the 1, 2, and 3 blocks from the left in the lower tier correspond to the matching chromaticity. In the upper leftmost block, the chromaticity at the upper left of the four points is determined to be a non-skin color pixel, and therefore a block including this is determined to be outside the skin color range. Similarly, the upper right 1 block and the lower right 2 block are out of range.

  Next, as a detailed explanation flowchart of the candidate group detection method in step S1001 of FIG. 10, a flowchart for explaining detection of the human skin area considered to be the most important in the image is shown in FIG. 6, and the details of step S1001 are further explained. To do.

  In step S601, DCT data and a quantization table of 8 * 8 pixel block units are acquired from the acquired Jpeg image data, and at the same time, the image data is developed into RGB bitmap data.

  Next, in step S602, the luminance and chromaticity data of each 8 * 8 pixel block constituting the image data are extracted, and classification is performed by comparing with the threshold values prepared in advance.

  The search range may be performed on the entire screen, but in this embodiment, since it is difficult to think that the human skin area of interest is located at the edge of the image, the area for several blocks from the edge of the image is The search is excluded from the chromaticity search target. In this case, since the ratio of the 8 * 8 pixel block image to the entire image differs depending on the input image size, the end portion proportional to the input image size is set. For example, it is 8 blocks for VGA (640 * 480) and 20 blocks for UXGA (1600 * 1200) images.

  Details of this portion will be described with reference to the flowchart of FIG.

  In step S201, luminance data is collected for each block obtained by dividing the input image into 8 * 8 pixel blocks. As a DC component (direct current component) in Jpeg luminance data, a luminance average value of 8 * 8 pixel data is detected as a representative value.

  Next, in step S202, chromaticity determination means is determined for each 8 * 8 pixel block to be subjected to chromaticity determination.

  In the chromaticity determination, there is a possibility of variation in the value of the RG chromaticity distribution of the skin color that is the matching range in the determination. In the variable, although not shown, in the internal processing of candidate group detection from step S1001 to step S1004, chromaticity determination is performed on the first detected candidate. Then, it is determined whether it is necessary to reset the conformity range for chromaticity determination, and when it is determined that remeasurement is necessary, the value of the RG chromaticity distribution of the skin color that is the conformity range for remeasurement is calculated. , Perform remeasurement using that value. Details will be described later.

  In step S203, chromaticity determination is executed for each 8 * 8 pixel block according to the determined chromaticity determination means.

  Next, in step S204, based on the determination result in step S203, the chromaticity of each 8 * 8 pixel block is selected as a feature region target. For the region of 8 * 8 pixel block selected by this selection, detection and determination of the feature region are advanced by the degree of coincidence with the definition range such as the size of the region and the feature amount by the spatial frequency characteristic. Details will be described later.

  By making the above detection block-based, a target having a specific unit is targeted, so that it is difficult to be affected by external noise. However, since the block of 8 * 8 pixels cannot be said to be an appropriate group size, even in the block detection based on chromaticity, it is further possible to perform detection with a restriction of continuous detection of blocks adjacent in the vertical and horizontal directions. The accuracy can be increased.

  Here, even if it is a person's skin color, when it prints, the thing which does not satisfy | fill the data amount which can recognize a face may be rejected as a non-adaptation, and the continuous range which determines such a thing as noise is set. This part is shown after step S603 in FIG. Hereinafter, a description will be given with reference to FIG.

  In step S603, detection is performed for each block determined to be suitable for the chromaticity range in the longitudinal direction with respect to the image, and candidates are searched in descending order of the number of consecutively detected blocks.

  Next, in step S604, it is determined whether a continuous detection block suitable for the suitable chromaticity range has been detected. If there is a target continuous detection block, the process proceeds to step S606. If there is no target continuous detection block, the process proceeds to step S605.

  In step S605, it is set that there is no feature area to be extracted in the input image data, and the process is terminated. Here, a value indicating that there is no feature region to be extracted is set in the predetermined identifier.

  On the other hand, in step S606, a block including a continuous amount defined in advance is searched for each block that is determined to be compatible with the appropriate chromaticity range in the short direction of the image, and the process proceeds to step S607.

  In step S607, it is determined whether there is a feature region candidate to be extracted in the input image data. If a non-existent result is obtained, that is, if there is no detected data, the process proceeds to step S605 to set that there is no feature area (similar to S605), and the process ends. On the other hand, if there is target candidate data, the process proceeds to step S608.

  In step S608, candidate numbers (1 to n) are assigned to feature region candidate data to be extracted in descending order of the number of continuous blocks in the longitudinal direction. When there is only one feature area candidate data, the number is 1, but when there are a plurality of feature area candidate data, up to 8 candidate numbers are issued. From among the areas having candidate numbers, a final target area is determined.

  Next, in step S609, it is confirmed that the candidate number of the feature area candidate data to be determined is not larger than the issued number (n), and the process proceeds to step S610. If it is greater than the candidate number, it is determined that the determination has been completed for all candidates, the process ends, and the process proceeds to step S1002 in FIG.

  In step S610, with respect to the feature region candidate data that is a determination target adapted to the chromaticity determination, whether the spatial frequency component is within the feature amount adaptation range (a range of spatial frequency characteristics set in advance) to be extracted. Make a comparison.

  In the Jpeg file, pixel information of all image regions is converted into spatial frequency components in groups of 8 * 8 pixels. The spatial frequency component referred to here is obtained by DCT-converting pixel data in units of 8 * 8 pixels for data compression of a Jpeg image. As is well known, Jpeg image data has an average value of 8 * 8 pixel data with the DC component (DC component) at the upper left of the 8 * 8 block data, and the other is AC component (AC component). The values are arranged in ascending order. That is, the spatial frequency data is arranged in order from the low frequency. By using the processed data as a frequency distribution and statistically processing the data in each block of the candidate group, it is determined whether the feature amounts in the spatial frequency component match. Note that the statistical processing here can use a general method for determining the coincidence of feature amounts.

  In step S611, a determination is made based on the result of step S610. If it is outside the conforming range, the process proceeds to step S609, and the process proceeds to determination for the next feature area candidate data.

  In step S612, the target feature region candidate data is recognized and set as a desired feature region. Then, the process proceeds to step S609, and the process proceeds to determination for the next feature area candidate data.

  The detection candidate having a higher priority (the detection block having a longer length) becomes a priority candidate, and adjacent chromaticity blocks are grouped. The groups of detection results are arranged in order of priority candidates, for example, in the order of color codes (1 = brown, 2 = red, 3 = orange, 4 = yellow, 5 = green, 6 = blue, 7 = purple, 8 = gray). The

  As described above, in the processing from step S1001 to step S1004, detection of a human skin region that is a desired feature region is executed. However, although it is not shown in the flowchart for the group of feature regions that are final candidates, it is determined whether the detection result can be used as it is. If it is determined that re-measurement is necessary, the value of the RG chromaticity distribution of the skin color that is the compatible range for re-measurement is calculated, and the re-measurement is performed by the processing from step S1001 to step S1004 using that value. Execute.

  FIG. 3 is a flowchart of a part for determining whether re-measurement by resetting is necessary and further processing. This will be described below.

  Step S301 represents the processing from step S1001 to step S1004 described above. Here, detection processing is performed to determine whether there is a feature region candidate group of a human skin region that is a desired feature region in the image.

  In step S302, if a feature area candidate group is detected as a result of the process, the process proceeds to step S303. On the other hand, if no feature region candidate group is detected, or if it is determined that the feature region candidate group is not a human skin region that is a desired feature region, the process proceeds to step S306. In step S306, the fact that the detection of the human skin area as the desired feature area has failed is set [HM = “0000”].

  In step S303, the representative chromaticity of the feature region candidate group is calculated. As the calculation of the representative chromaticity, an average value of the detected chromaticities in block units constituting the feature region candidate group is calculated, and this is used as the representative value.

  Next, in step S304, the representative chromaticity, which is the chromaticity representative value of the feature region candidate group, is determined based on the degree of separation from the center of the chromaticity range used for the chromaticity suitability determination, and the skin color of the feature region candidate group It is determined whether or not the chromaticity range determination value is effective for area detection. Details will be described later. If it is determined that the human skin area, the process proceeds to step S305.

  In step S305, the fact that the human skin area has been successfully detected is set [HM = “1111”]. Further, feature amount data of the detected feature region is stored.

  Here, determination of an effective chromaticity range will be described.

  FIG. 4 shows actual data obtained by measuring and plotting the chromaticity of a human skin area as a subject from a plurality of unspecified image data. The vertical axis represents the chromaticity (G / R + G + B) of the green component. The horizontal axis represents the chromaticity (R / R + G + B) of the red component.

  A conventional measurement method that makes a determination in a single measurement with a fixed chromaticity range is plotted with a circle (○), and detection has failed or has been detected. The case where the completed area is incomplete is plotted with (Δ). In many cases, the detection has succeeded in the range that falls within the matching chromaticity shown in the above equation (4), and the skin color of the subject distributed near or outside the boundary of the matching chromaticity range has failed to be detected. I understand that. The cause of this is considered to be a color cast as a shooting environment. Even in this case, for those near the boundary of the adaptive chromaticity range, by resetting the adaptive chromaticity represented by the equation (4) based on the chromaticity representative value of the detected feature region candidate group, Good detection is possible.

  Next, a method for determining whether the determination value of the chromaticity range is effective for detecting the skin color area of the feature area candidate group will be described.

  The reference used in the chromaticity determination in step S302 is the chromaticity range expressed by the above-described equation (4). That is, the compatible range of the chromaticity [R / (R + G + B)] of the red component is “0.35 to 0.44”, and the compatible range of the chromaticity [G / (R + G + B)] of the green component is “0.29 to 0. .33 ″ where the skin color region of the feature region candidate group falls is determined as the human skin region.

  FIG. 8 shows a detection result using the conventional detection method. In FIG. 8, 1) is an original image with a person as a subject. 8) is a feature region candidate group in the case where the skin color region conformity determination is determined based only on the criterion of equation (4). Comparing the upper left detection area (part corresponding to the upper left human face) that is the first candidate group with the person's skin area in the actual original image, it can be seen that the detected area is less than the actual area. .

  Among the feature region candidate groups, the chromaticity representative value in the first candidate group is that the red component chromaticity [R / (R + G + B)] is “0.375”, and the green component chromaticity [G / (R + G + B)]. The matching range was calculated as “0.300”.

  FIG. 13 shows the relationship between the compatible chromaticity range defined by the equation (4) and the chromaticity representative value in the first candidate group of FIG.

  In the figure, the vertical axis represents the chromaticity (G / R + G + B) of the green component, with the lowest value being “0.290” and the highest value being “0.330”. The horizontal axis represents the chromaticity (R / R + G + B) of the red component, where the lowest value is “0.350” and the highest value is “0.440”. In the center “O”, the chromaticity of the green component is “0.310” and the chromaticity of the red component is “0.395”. The region is divided into a region in which the radius “R” is within the range of “0.020” with the center “O” as the center and the other regions.

  For the case where the chromaticity representative value in the first candidate group falls within the radius “R” with the center “O” as the center, the determination value of the chromaticity range that is effective for the skin color area detection in step S304 judge.

  In addition, when the representative chromaticity value in the first candidate group is not included in the radius “R” with the center “O” as the center, the chromaticity range is defined in the equation (4). If so: That is, at the detection position of the first candidate group in FIG. 8 (marked with a star in FIG. 13), the detection color that falls within the applicable range as the human skin color, but proceeds to step S307 in step S304 in the flow of FIG. It is determined that the degree range needs to be reset and reprocessed.

  In step S307, with the chromaticity representative value in the first candidate group as the center, the adaptation width “0.090” of the chromaticity of the red component (R / R + G + B) and the adaptation width of the chromaticity of the green component (G / R + G + B) Calculate and set a new chromaticity matching criterion for detection with “0.040”. In this example, the width of the adaptive chromaticity range is the same as the initial value, but the center is different.

  FIG. 9 shows a result of creating a new detection chromaticity matching criterion centering on the chromaticity representative value in the first candidate group detected from the original image of FIG. 8 in this case. The chromaticity range, which is the detection criterion for the initial value, is indicated by a solid line, and the chromaticity range, which is the adaptation criterion for the new detection chromaticity, is indicated by a dotted line. It can be confirmed that the adaptive chromaticity range has moved.

  Thereafter, in step S308, detection processing equivalent to that in step S301 is executed.

Next, in step S309, if it is determined that no feature region candidate group has been detected, or if it is determined that the feature region candidate group is not a human skin region that is a desired feature region, the process proceeds to step S306.
In step S306, the fact that the detection of the human skin area as the desired feature area has failed is set [HM = “0000”].

On the other hand, if a feature region candidate group is detected in step S309, the process proceeds to step S305.
In step S305, the fact that the human skin area has been successfully detected is set [HM = “1111”]. Then, the feature amount data of the detected feature region is stored.

  3) of FIG. 8 shows the result of detecting a desired feature region using the above-described new detection chromaticity matching criteria. It can be confirmed that the detection is clearly better than the detection result of 2) in FIG.

  In the description of the present embodiment, the determination for remeasurement using the representative value of the final candidate group and the value of the RG chromaticity distribution of the skin color that is the matching range for remeasurement are calculated. At this time, when there are a plurality of final candidate groups, it is also more effective to calculate the value of the RG chromaticity distribution of the skin color that is the compatible chromaticity range for remeasurement using the representative value for each of the plurality of groups. It is.

  In the present embodiment, a method using a “JPEG file” as an image compression file has been disclosed. Of course, it is needless to say that the feature region detection can be realized by a simple process with respect to other files using the conversion to the frequency component such as “Jpeg2000 file” in the same way.

  The image processing apparatus according to the present embodiment described above is constituted by a computer CPU or MPU, RAM, ROM, etc. as described above, and the program stored in the RAM or ROM operates. realizable.

  Therefore, a program that causes a computer to perform the above functions can be realized by recording the program on a recording medium such as a CD-ROM and causing the computer to read the program. As a recording medium for recording the program, a flexible disk, a hard disk, a magnetic tape, a magneto-optical disk, a nonvolatile memory card, and the like can be used in addition to the CD-ROM.

  A program in which the functions of the above-described embodiments are realized by executing a program supplied by a computer is included in the embodiments of the present invention. Such a program may realize the functions of the above-described embodiment in cooperation with an OS (operating system) running on a computer or other application software. Alternatively, all or part of the processing of the program supplied to the computer may be performed by the function expansion board or function expansion unit of the computer to realize the functions of the above-described embodiment.

  In order to use the present invention in a network environment, all or a part of the program may be executed on another computer. For example, the screen input process may be performed by a remote terminal computer, and various determinations, recording, and the like may be performed by another center computer.

It is a flowchart showing the flow of the whole process in connection with one Embodiment of this invention. It is a flowchart of the brightness | luminance and chromaticity detection and determination of a pixel group concerning one Embodiment of this invention. 6 is a flowchart of redefinition determination and processing in chromaticity definition according to an embodiment of the present invention. It is the data of the chromaticity distribution in the area | region which was extracted from the several image sample concerning one Embodiment of this invention, and the detection is inadequate. It is RG chromaticity distribution of the skin color in connection with one Embodiment of this invention. It is a flowchart of the person skin area | region detection from the JPEG image decompression | decompression concerning one Embodiment of this invention. It is an example of the chromaticity determination method in 8 * 8Block which is a JPEG file image compression unit according to an embodiment of the present invention. It is a figure which shows (1) the original image which is a target image, (2) the conventional detection result (chromaticity 1pass), (3) the detection result (chromaticity 2pass) by this invention. It is a figure which shows the example at the time of recalculating the detection chromaticity range with respect to the suitable chromaticity range in connection with one Embodiment of this invention. It is a general | schematic flowchart of the person skin area | region detection process in connection with one Embodiment of this invention. 1 is a hardware configuration of an image processing apparatus according to an embodiment of the present invention. It is a figure which shows the example of the establishment ellipse range which shows RG chromaticity distribution of the skin color in connection with one Embodiment of this invention. It is a figure showing the relationship between the RG chromaticity range of the skin color in connection with one Embodiment of this invention, and the chromaticity representative value in a 1st candidate group.

Explanation of symbols

51 ... Input / output port 52 ... Mouse 53 ... Keyboard 54 ... Monitor 55 ... Control unit 56 ... Memory card drive 57 ... Hard disk

Claims (12)

In an image processing apparatus for detecting a specific feature region from image data,
First detection means for detecting a candidate for the feature region based on a preset chromaticity range;
Resetting means for resetting the chromaticity range based on chromaticity information of candidate feature regions detected by the first detection means;
Second detection means for further detecting the candidate for the feature region based on the chromaticity range reset by the resetting means;
An image processing apparatus comprising: a feature region determination unit that determines whether a candidate for a feature region detected by the first detection unit or the second detection unit is the specific feature region. In an image processing apparatus for detecting a specific feature region from image data,
First detection means for detecting a candidate for the feature region based on a preset chromaticity range;
Resetting determination means for determining whether or not the chromaticity range needs to be reset based on chromaticity information of candidate feature regions detected by the first detection means;
When the reset determination unit determines that the chromaticity range needs to be reset, the chromaticity range is reset based on the chromaticity information of the feature region candidates detected by the first detection unit. Resetting means to
Second detection means for further detecting feature region candidates based on the chromaticity range reset by the resetting means;
An image processing apparatus comprising: a feature region determination unit that determines whether a feature region candidate detected by the first and second detection units is the specific feature region.   The image processing apparatus according to claim 1, wherein the detection processing in the first and second detection units is performed on the image data in units of a block having a specific size.   The image processing apparatus according to claim 3, wherein the feature region candidates include a group of consecutive blocks detected as candidates.   The image processing apparatus according to claim 1, wherein the resetting unit resets the chromaticity range for each feature region candidate.   The image processing apparatus according to claim 1, wherein the feature region determination unit performs the determination based on a spatial frequency characteristic of the feature region candidate.   The image processing apparatus according to claim 1, wherein the feature region includes at least a human skin region.   The image processing apparatus further includes an image correction unit and a correction information generation unit, wherein the correction information generation unit generates correction information for the feature region, and the image correction unit uses the correction information generated for the feature region to generate an image. 8. The image processing apparatus according to claim 1, wherein correction processing is performed, and image correction processing defined in advance for all regions is performed for other regions. In an image processing method for detecting a specific feature region from image data,
A first detection step of detecting a candidate for the feature region based on a preset chromaticity range;
A resetting step of resetting the chromaticity range based on the chromaticity information of the feature region candidates detected in the first detection step;
A second detection step for further detecting the feature region candidates based on the chromaticity range reset in the resetting step;
An image processing method comprising: a feature region determination step of determining whether a feature region candidate detected in the first detection step or the second detection step is the specific feature region. In an image processing method for detecting a specific feature region from image data,
A first detection step of detecting a candidate for the feature region based on a preset chromaticity range;
A reset determination step for determining whether or not it is necessary to reset the chromaticity range based on chromaticity information of candidate feature regions detected in the first detection step;
When it is determined in the reset determination step that the chromaticity range needs to be reset, the chromaticity range is reset based on the chromaticity information of the feature region candidates detected in the first detection step. A reconfiguration step to
A second detection step of further detecting feature region candidates based on the chromaticity range reset in the resetting step;
An image processing method comprising: a feature region determination step of determining whether the feature region candidates detected in the first and second detection steps are the specific feature region.   A non-transitory computer-readable storage medium storing a program for causing a computer to realize the function of each unit included in the image processing apparatus according to claim 1. A computer-readable recording medium on which the program according to claim 11 is recorded.

Images