JP2009064459A - Program for correcting red eye in image, recording medium and pink eye correction method - Google Patents

Abstract

<P>PROBLEM TO BE SOLVED: To appropriately correct red eyes in an image, while considering gradation. <P>SOLUTION: A plurality of feature areas requiring red eye correction are extracted from an image on the basis of saturation, lightness and hue that are feature quantities, and a histogram is generated for each of the feature quantities of the feature areas (a step S32). Parameters such as a minimum, a maximum, a mode, a variance, an average and so on of the feature quantities are derived from the histograms, and a correction curve for the feature quantities is generated on the basis of the parameters (steps S33, S34). The feature quantities are corrected according to the correction curve (a step S35). As a result of this, the feature quantities of the feature areas can be appropriately corrected by using the correction curve for each of the feature quantities according to the property of the feature areas, and red eyes occurring in various aspects can be appropriately corrected, while considering gradation. <P>COPYRIGHT: (C)2009,JPO&INPIT

Description

  The present invention relates to a technique for correcting red eyes in an image.

  When shooting with a flash, there may be a red-eye phenomenon in which the eyes of a person in the image shine in red or golden. Conventionally, the image data acquired by taking a photograph acquired with a silver salt camera with a scanner or taking a picture with a digital camera is processed, and the red eyes in the image are A technique for correcting the error has been proposed.

  As a general technique for correcting red-eye, a process of replacing red with another color in an area where red-eye exists in an image (hereinafter referred to as “red-eye area”), or reducing brightness or saturation. It has been known. When such a simple process is applied to the red-eye region, the saturation, lightness, and hue gradation in the eye are lost, resulting in an unnatural eye. Therefore, for example, in Patent Document 1 and Patent Document 2, brightness or saturation is gradually decreased from the peripheral part to the central part of the pupil or red-eye region.

JP 2000-76427 A JP 2000-134486 A

  By the way, when correcting the lightness and saturation in the red-eye region in accordance with the position of the pixel to be corrected, it is necessary to appropriately associate the position of the correction target pixel and the correction amount. Therefore, for example, when the center of the pupil cannot be detected properly, or when the center of the red-eye region is regarded as the center of the pupil even though the center of the red-eye region does not coincide with the center of the pupil, the gradation near the center of the pupil is used. Will be unnatural.

  The present invention has been made in view of the above problems, and an object thereof is to appropriately correct red eyes generated in various modes.

  The invention according to claim 1 is a program for correcting red-eye in an image, and the execution of the program by a computer causes the computer to extract a plurality of feature regions extracted as red-eye regions from an image based on feature amounts. And a step of performing smoothing according to the feature region to be processed on each of the plurality of feature regions before or after the step of performing the red eye correction.

  Invention of Claim 2 is a program of Claim 1, Comprising: Execution by the said computer of the said program produces | generates the histogram of the feature-value regarding each of these feature area in the said computer, And a step of determining the degree of smoothing based on a parameter derived from the histogram.

  The invention according to claim 3 is a program for correcting red-eye in an image, and the execution of the program by a computer determines a correction target area in the image to the computer, and A step of comparing the size with a predetermined value, a step of increasing the resolution of the correction target region when the size of the correction target region is smaller than the predetermined value, and red-eye correction on the correction target region. And performing the process.

  The invention according to claim 4 is a computer-readable recording medium in which the program according to any one of claims 1 to 3 is recorded.

  The invention according to claim 5 is a red-eye correction method for correcting red eyes in an image, the step of performing red-eye correction on a plurality of feature regions extracted as red-eye regions from an image based on feature amounts; Before or after the step of performing the red-eye correction, a step of performing smoothing according to the feature region to be processed on each of the plurality of feature regions.

  The invention according to claim 6 is a red-eye correction method for correcting red-eye in an image, the step of determining a correction target region in the image, and the step of comparing the size of the correction target region with a predetermined value. When the size of the correction target area is smaller than the predetermined value, the method includes a step of increasing the resolution of the correction target area and a step of performing red-eye correction on the correction target area.

  According to the first, second, and fifth aspects of the invention, it is possible to perform appropriate smoothing on the entire red-eye region by performing smoothing according to the feature region.

  Further, according to the invention of claim 2, a parameter for determining the degree of smoothing can be easily prepared.

  According to the third and sixth aspects of the invention, the accuracy of red-eye correction can be improved.

  Embodiments of the present invention will be described below with reference to the drawings.

<Embodiment>
FIG. 1 is an external view of an image processing apparatus 1 according to an embodiment of the present invention. The image processing apparatus 1 is a computer that performs correction by specifying a red-eye region in an image by executing a program. As shown in FIG. 1, the image processing apparatus 1 includes a keyboard 111 and a mouse 112 that receive input from the user, and a display 12 that displays an instruction menu for the user, an acquired image, and the like.

  The image processing apparatus 1 includes a fixed disk 161 that stores therein image data and the like. Further, a recording disk 91 that stores a program and a memory card 92 that stores image data are computer-readable recording media. Can be loaded into the reader 162 and the card slot 163, respectively.

  FIG. 2 is a block diagram illustrating a configuration of the image processing apparatus 1. The image processing apparatus 1 has a general computer system configuration in which a CPU 13, a RAM 14, and a ROM 15 are connected to a bus line. The bus line further includes a display 12, a keyboard 111 and a mouse 112 for receiving input from an operator, a fixed disk 161 for storing data and programs, and a recording disk 91 (optical disk, magnetic disk, magneto-optical disk, etc.). And a card slot 163 for exchanging information with the memory card 92 are appropriately connected via an interface (I / F) or the like.

  The RAM 14, the fixed disk 161, the reading device 162, and the card slot 163 can exchange data with each other. Under the control of the CPU 13, the display 12 stores various information and the RAM 14, the fixed disk 161, the memory card 92, and the like. The displayed image can be displayed.

  The program 141 shown in FIG. 2 is stored in the fixed disk 161 from the recording disk 91 via the reading device 162, transferred from the fixed disk 161 to the RAM 14, and can be executed by the CPU 13.

  FIG. 3 is a diagram showing a functional configuration realized by the CPU 13 operating according to the program 141 in the RAM 14 together with other configurations. In the configuration illustrated in FIG. 3, the display control unit 201, the target region determination unit 202, the size determination unit 203, the feature region extraction unit 204, the red-eye region specification unit 205, and the correction unit 206 indicate functions realized by the CPU 13 and the like. .

  The display control unit 201 controls display of an image on the display 12 based on the image data. The target area determination unit 202 receives from the user the determination of the target area to be processed in the image indicated by the image data 301 via the mouse 112. The size determination unit 203 determines the size of a unit area that is a unit for performing processing to be described later on the target area.

  The feature region extraction unit 204 performs processing for each unit region on the target region, and extracts a plurality of types of feature regions according to the feature amount. Note that saturation, brightness, and hue are used as the feature amount. The red-eye area specifying unit 205 specifies a feature area constituting the red-eye area from among the extracted feature areas. The correction unit 206 performs color conversion on the specified red-eye region, and corrects red eyes in the image to normal color eyes. The corrected image data is stored in the RAM 14 as corrected image data 302.

  FIG. 4 is a block diagram illustrating a functional configuration of the correction unit 206. The correction unit 206 includes a histogram generation unit 211 that generates a histogram of feature amounts in a feature region that constitutes a red-eye region, a correction curve generation unit 212 that generates a correction curve for correcting red eyes, and a feature amount using the correction curve. It has the conversion part 213 which converts, the filter production | generation part 214 which produces | generates the filter for smoothing, and the smoothing part 215 which smoothes the red eye area | region after correction | amendment. Details of these functions will be described later.

  5 to 7 are diagrams showing the flow of operations when the image processing apparatus 1 specifies and corrects the red-eye area. The operation of the image processing apparatus 1 for correcting the red eye (hereinafter referred to as “red-eye correction”) will be described below with reference to FIGS. 3 to 7.

  First, when the user selects a desired image data in the memory card 92 or the fixed disk 161 using the keyboard 111 or the mouse 112 while looking at the display 12, the selected image data 301 is read into the RAM 14. The display control unit 201 displays an image on the display 12 based on the image data 301 (step S11).

  The user uses the mouse 112 to specify an area that needs red-eye correction, so that the target area determination unit 202 determines a target area to be calculated (step S12). Specifically, as shown in FIG. 8, when the user designates two diagonal points of the target area 402 in the image 401, the target area 402 including the red-eye area corresponding to one eye is determined. .

  When the target region 402 is determined, the size determining unit 203 uses the number of pixels N1 of the target region 402 and the parameter S to determine the number of pixels on one side of the unit region (hereinafter referred to as “unit region size”). ) N (the size of the unit area is n × n) is obtained (step S13).

  The parameter S is a value indicating the ratio of the number of pixels between the image 401 and the target area 402, and is a parameter determined by the number 2 when the number of pixels of the image 401 is N2.

  The function F is a function that decreases the unit area size n to be output as the parameter S is small, and increases the unit area size n to be output as the number of pixels N1 of the target area 402 is large. Is a function having an input / output relationship such as

  By determining the unit region size using such a function F, the target region 402 can be divided into unit regions having a size corresponding to the ratio of the size of the target region 402 to the image 401. Thereafter, the unit region is divided into units. As a result, it is possible to specify a red-eye region with high accuracy and high accuracy that is less susceptible to noise than in the case of calculating for each pixel. Note that the minimum value of n is 1, in which case one unit area is one pixel, and the target area 402 is processed for each pixel.

  When the unit region size n is determined, the feature region extraction unit 204 converts the RGB value (average RGB value) of each unit region of the target region 402 into the color space of the HSL color system, and Hue, saturation and lightness are obtained as feature quantities (step S14).

  Next, the feature region extraction unit 204 extracts the pupil region, the iris region, and the non-iris region from the target region 402 as feature regions based on the feature amount of the unit region. Specifically, an area where saturation and brightness are in a certain range (area where both saturation and brightness are low) is extracted as a pupil area, and an area where hue and saturation are in a certain range (hue is red) (A region that is slightly yellow and has a high saturation range) is extracted as an iris region. Further, a region having the same brightness range as the pupil region and the same hue range as the iris region is extracted as an out-iris region (step S15).

  It should be noted that by extracting the feature region to be corrected based on the feature amount, it is realized that correction based on the feature amount described later is appropriately performed for each feature region.

  FIG. 9 is a diagram illustrating a feature region extracted from the target region 402 by the feature region extraction unit 204. Reference numeral 501 denotes a pupil area, reference numerals 502a and 502b denote iris areas, and reference numeral 503 denotes a non-iris area. Depending on the state of the red eye, various feature regions may not be extracted or a plurality of feature regions may be extracted.

  The range of the feature amount used for extracting various feature regions is not determined from the normal feature amount where the red-eye phenomenon does not occur, but is statistically determined from the feature amount when the red-eye phenomenon occurs. Therefore, it means that each feature region corresponding to such a feature amount is extracted as a region where the red-eye phenomenon may occur. Conversely, a region that is not extracted as any feature region is determined as a region that does not require red-eye correction.

  Next, unnecessary regions that are not suitable for red-eye correction are excluded from the extracted feature regions (step S16). For example, the minute iris region 502b in the feature region shown in FIG. 9 is excluded from the correction target as an erroneously extracted iris region. Further, based on the inclusion relationship of each feature region and the degree of contact, it is determined again whether each feature region is appropriate as a region to be corrected.

  As a specific example, if the extra-iris area is not in contact with most of the periphery of the iris area, correcting the extra-iris area results in unnatural eyes, so the extra-iris area should not be corrected. It is determined. If there is a region other than the feature region between the pupil region and the iris region, the periphery of the pupil region may not be red-eye, and it is determined that the pupil region is a region that should not be corrected. When the pupil region is not completely included in the iris region, it is determined that it is difficult to perform appropriate correction, so that no feature region should be corrected (in this case, red-eye correction is not performed). .

  When at least one feature region constituting the red-eye region to be corrected for red-eye is determined, information indicating the feature region and feature amount are input to the correction unit 206 from the red-eye region specifying unit 205 to configure the red-eye region. Processing for red-eye correction is sequentially executed for each feature region. The following processing is executed without depending on the type of feature region.

  First, the correction unit 206 determines one feature region included in the red-eye region as a processing target (hereinafter referred to as “target feature region”) (step S21). Then, the histogram generation unit 211 (see FIG. 4) counts the number of unit areas constituting the target feature area, and obtains the size of the target feature area as the count number of the unit area.

  When the size of the feature region of interest is equal to or smaller than the predetermined value (step S22), the fact is transmitted from the histogram generation unit 211 to the size determination unit 203, and the size determination unit 203 sets the unit region size to be smaller. The resolution is increased and the process returns to step S14 (step S23). As a result, the number of unit regions in the feature region extracted again is increased, and the reliability of parameters obtained from a histogram described later can be improved. As a result, the accuracy of red-eye correction can be improved.

  When one unit area is one pixel, the size determination unit 203 performs a process for increasing the resolution on the target area 402 (or the target feature area). As processing for improving the resolution (that is, processing for enlarging an image), linear interpolation, cubic convolution interpolation, B-spline interpolation, or the like is used. As a result, the number of pixels in the feature region extracted again is increased.

  Next, the histogram generation unit 211 determines a feature amount (one of saturation, lightness, and hue, hereinafter referred to as “target feature amount”) to be processed in the target feature region (FIG. 6: step S31). ), Correction of the feature of interest (steps S32 to S35) and smoothing of the feature of interest (steps S42 and S43).

  In the correction of the feature of interest, the histogram generator 211 generates a histogram of the feature of interest in the feature region of interest (step S32). That is, a histogram indicating the relationship between the value of the feature amount of interest and the number of unit regions having the value is generated.

  When the histogram of the feature amount of interest with the unit area as a unit is generated, a parameter that is a statistical value derived from the histogram is obtained by the correction curve generation unit 212 (step S33). A correction curve indicating the relationship between the value before correction of the feature value of interest and the value after correction is generated (step S34). Then, the attention feature amount of each unit region is converted by the conversion unit 213 using the correction curve (step S35). Thereby, the feature amount of interest in the feature region of interest is corrected.

  The smoothing of the attention feature amount after correction is executed only when the attention feature amount is saturation or lightness (step S41), and the variation in saturation and lightness emphasized by the correction (variation in the two-dimensional space) is smoothed. Is alleviated by In smoothing, a weighting coefficient for smoothing is obtained from the histogram (step S42), and the target feature amount of the unit area is corrected to a smoothed value (step S43). Thereby, smoothing is performed at an appropriate degree for each feature region.

  When correction and smoothing are completed for one target feature amount, the next target feature amount is determined (steps S44 and S31), and correction and smoothing are performed again. In the present embodiment, since the smoothing for the hue is omitted, the saturation and the lightness are corrected and smoothed, and the hue is only corrected. However, the hue may be smoothed.

  Further, in FIG. 6, smoothing is performed after correction, but smoothing may be performed before correction. In this case, steps S41 to S43 and steps S34 and S35 are interchanged. Even when smoothing is performed before correction, variations in saturation and brightness emphasized by the correction are alleviated. Further, by performing smoothing before correction, noise in the feature region can be reduced before correction.

  Next, the correction curve and smoothing used for correction will be described in more detail.

  FIG. 10 is a diagram illustrating a correction curve 61 relating to saturation. In FIG. 10, the horizontal axis represents the saturation before correction (that is, the input of the conversion unit 213), and the vertical axis represents the saturation after correction (that is, the output of the conversion unit 213). The value XS1 is the minimum saturation value before correction in the target feature region, and the value XS3 is the maximum saturation value before correction. The value XS2 is the mode of saturation (the saturation of the most unit areas). These values are derived from the saturation histogram.

  In FIG. 10, the corrected values YS1, YS2, and YS3 correspond to the values XS1, XS2, and XS3, respectively. The value YS1, which is the minimum saturation, is made equal to the value XS1. Thereby, it is possible to prevent the saturation of the red-eye area from being excessively reduced in the corrected image.

  The value YS3 is obtained by the function FS3 expressed by Equation 6.

  The function FS3 uses the maximum value XS3 and the saturation variance VS obtained from the saturation histogram as parameters, and has a characteristic that the value YS3 decreases as the variance VS decreases. That is, the function FS3 obtains the value YS3 by subtracting the larger value ZS3 from the maximum value XS3 as the variance VS is smaller.

  In red-eye correction, it is preferable that the maximum value of saturation is kept as small as possible. On the other hand, if the maximum value of saturation is reduced when the variance VS is large, there is a problem that the atmosphere of the feature region of interest changes. Therefore, the saturation correction curve 61 has a characteristic of decreasing the maximum value of saturation as the variance of saturation is smaller by the function FS3.

  Since the value YS3 is not allowed to be less than or equal to the value YS1, the function FS3 uses the value YS1 (that is, the minimum value XS1) as an auxiliary parameter.

  On the other hand, the value YS2 is obtained by the function FS2 expressed by Equation 7. It should be noted that the degree of decrease in the intermediate value of saturation (which is an approximately intermediate value in the saturation range; the same applies hereinafter) is determined based on the mode value XS2 in the vicinity of the mode value XS2. This is because the purpose of red-eye correction, which reduces the saturation, can be easily and accurately realized by greatly reducing.

  The function FS2 uses the mode value XS2 and the variance VSM in the vicinity of the mode value in the saturation histogram as parameters. The smaller the variance VSM, the smaller the value YS2. That is, the function FS2 obtains a value YS2 by subtracting a larger value ZS2 from the mode value XS2 as the variance VSM is smaller. Since the value YS2 needs to be a value between the value YS3 and the value YS1, the function FS3 uses the value YS3 (or the maximum value XS3 and the variance VS for obtaining the value YS3) and the value YS1 (that is, the minimum The value XS1) is used as an auxiliary parameter.

  When the variance VSM in the vicinity of the mode value is small, the saturation histogram is concentrated in the vicinity of the mode value XS2. Therefore, by setting the value YS2 to a small value when the variance VSM is small, it is possible to appropriately obtain the red-eye correction effect of reducing the saturation. On the other hand, when the variance VSM is large, the saturation histogram is not so biased to the mode value XS2, and therefore the effect of efficiently reducing the saturation cannot be expected even if the value YS2 is small. Rather, it will make the saturation unevenness stand out. Therefore, in the function FS2, the value YS2 is obtained by subtracting a larger value from the value XS2 as the variance VSM is smaller.

  Note that the value YS2 may be adjusted based on the difference between the maximum value XS3 and the mode value XS2 by the function FS2. For example, when the difference between the maximum value XS3 and the mode value XS2 is small, the value YS2 may be slightly increased so that the correction curve 61 does not bend excessively.

  When the values YS1, YS2, and YS3 are determined as described above, a quadratic curve that passes through the three points of coordinates (XS1, YS1), (XS2, YS2), and (XS3, YS3) is a correction curve related to saturation. 61 is generated. The correction curve 61 may be obtained from the three points by another method.

  FIG. 11 is a diagram illustrating a correction curve 62 relating to brightness. In FIG. 11, the horizontal axis represents the lightness before correction (that is, the input of the conversion unit 213), and the vertical axis represents the lightness after correction (that is, the output of the conversion unit 213). The value XL1 is the minimum value of brightness before correction, and the value XL3 is the maximum value of brightness before correction. The value XL2 is the mode value of brightness. These values are derived from the brightness histogram.

  In FIG. 11, similarly to FIG. 10, the corrected values YL1, YL2, and YL3 correspond to the values XL1, XL2, and XL3, respectively. The value YL1 which is the minimum lightness after correction is made equal to the value XL1, and the value YL3 which is the maximum lightness is made equal to the value XL3. The reason why the value YL1 is made equal to the value XL1 is to prevent the red-eye region from being crushed black in the corrected image and resulting in an unnatural eye, and the value YL3 and the value XL3 are made equal. This is to make the corrected eye alive by preventing the disappearance of the so-called catchlight (the area where the illumination is reflected and brightened) existing in the eye.

  The value YL2 is obtained by the function FL2 expressed by Equation 8. Note that the degree of decrease in the intermediate value of lightness is determined based on the mode value XL2, as in the case of saturation, the lightness is decreased by greatly reducing the lightness in the vicinity of the mode value XL2. This is to easily and accurately realize the purpose of red-eye correction.

  The function FL2 uses the mode value XL2 and the variance VL of the brightness histogram as parameters. The smaller the variance VL, the smaller the value YS2. That is, the function FL2 obtains the value YL2 by subtracting the larger value ZL2 from the mode value XL2 as the variance VL is smaller. Since the value YL2 needs to be greater than or equal to the value YL1, the function FL2 also uses the value YL1 (that is, the minimum value XL1) as an auxiliary parameter.

  When the variance VL is large, various brightness values exist in the feature region of interest, and it is assumed that there is a strong gradation related to the brightness value. As a typical example, a state in which a plurality of small regions having different brightness values exist in the feature region of interest can be considered. In such a case, if the value YL2 is set to a small value, the boundary of the small areas with different brightness in the attention feature area after correction becomes clear, and the unevenness of brightness becomes conspicuous.

  Therefore, the function FL2 is a function that decreases the value YL2 only when the variance VL is small. As a result, it is possible to reduce the brightness as much as possible while preventing a pattern such as a step in the feature region from being generated by red-eye correction.

  In the function FL2, as in the case of saturation, the value YL2 may be adjusted based on the magnitude of the difference between the maximum value XL3 and the mode value XL2.

  When the values YL1, YL2, and YL3 are determined as described above, a quadratic curve that passes through the three points of coordinates (XL1, YL1), (XL2, YL2), and (XL3, YL3) is a correction curve 62 relating to brightness. Is generated as As a result, a correction curve 62 that decreases the intermediate value while maintaining the minimum value and the maximum value of the brightness is obtained. Of course, the correction curve 62 may be obtained from the three points by another method.

  FIG. 12 is a diagram illustrating a correction curve 63 related to hue. In FIG. 12, the horizontal axis represents the hue before correction (that is, the input of the conversion unit 213), and the vertical axis represents the hue after correction (that is, the output of the conversion unit 213). The value XH2 is the mode value of the hue histogram, and the range between the value XH1 and the value XH3 is a hue range that can be regarded as the same color as the hue indicated by the value XH2. However, all the hues in the target feature area are included in the range between the value XH1 and the value XH3. In FIG. 12, the corrected values YH1, YH2, and YH3 correspond to the values XH1, XH2, and XH3, respectively.

  Note that the value XH1 and the value XH3 may be determined as values at one end and the other end of the hue range before correction in the target feature region.

  The corrected values YH1, YH2, and YH3 are set in advance, and are set as values corresponding to the brown range in the case of a brown iris and to the blue range in the case of a blue iris. Then, a quadratic curve passing through the three points of coordinates (XH1, YH1), (XH2, YH2), and (XH3, YH3) is generated as a correction curve 63 related to hue. Of course, the correction curve 63 may be obtained from the three points by another method.

  By generating the correction curve 63 by the above method, the hue of the feature region of interest is corrected to a value within an appropriate hue range while correcting the mode value XH2 before correction to the value YH2 specified in advance. Is realized.

Next, the calculation and smoothing of the weighting coefficient for smoothing performed on the saturation and the brightness will be described. Smoothing is a process of converting the feature quantity of the unit area to be processed into a value obtained by weighted averaging together with the feature quantities of the surrounding eight unit areas. For example, if any feature amount of 3 × 3 (= 9) unit areas is C _p (p indicates the position of the unit area) and the weighting coefficient for this unit area is a _p , smoothing is performed. The subsequent feature amount C is obtained by Equation 9.

  FIG. 13 is a diagram showing filter weighting coefficients used in smoothing. As shown in FIG. 13, the weighting coefficient of the central unit area to be processed is set to 1, and the weighting coefficient of the surrounding unit areas is set to 1 / W. However, the value W is a value of 1 or more, and is obtained from Equation 10 when the feature amount of interest is saturation and is obtained from Equation 11 when the feature value is lightness.

  The saturation function FWS uses the saturation mode XS2 and the variance VS as parameters, and these parameters are obtained from the saturation histogram. The lightness function FWL also uses the lightness mode XL2 and the variance VL as parameters, and these parameters are obtained from the lightness histogram.

  As described above, when the variance of the feature amount is large, there is a high possibility that a plurality of small regions having different feature amounts exist in the feature region of interest. When red-eye correction is performed in such a case, the difference in saturation and lightness for each small area becomes large, and unevenness occurs in the feature area of interest. Therefore, the functions WWS and FWL determine the value W so that the degree of smoothing increases as the variances VS and VL increase. Note that the degree of smoothing increases as the value W approaches 1.

  Further, when there are a plurality of small regions having different feature amounts in the feature region of interest, the bias in saturation and lightness after correction becomes more prominent as the mode values XS2 and XL2 of saturation and lightness are larger. For FWS and FWL, the value W is determined so that the smoothing degree increases as the mode values XS2 and XL2 increase.

  As described above, by determining the degree of smoothing based on the variance and the mode value, appropriate smoothing according to the characteristics of each feature region is realized.

  When the correction for the target feature region and the necessary smoothing are completed, it is confirmed whether or not there is an unprocessed feature region (FIG. 7: Step S51). A feature region is determined (FIG. 5: step S21). Thereafter, correction and smoothing are performed on the saturation and lightness of each unit region of the target feature region, and the hue is corrected.

  When correction and smoothing are performed on all feature amount regions, the target region 402 having the corrected feature region is returned from the HSL color space to the RGB color space, and further combined with the image data 301 to be corrected. The image data 302 is stored in the RAM 14 (step S52). The corrected image data 302 is transferred to the display control unit 201, and an image subjected to red-eye correction is displayed on the display 12 (step S53).

  The image processing apparatus 1 that performs red-eye correction according to the program 141 has been described above. However, the image processing apparatus 1 generates a histogram of feature amounts in a feature region that is a correction target, and a correction curve for correcting the feature amounts is displayed in the histogram. Based on. Accordingly, it is possible to perform correction according to the characteristics of the feature region with respect to the feature region. In addition, since gradation is generated in the corrected feature area in accordance with the gradation of the feature amount in the feature area before correction, correction can be performed to more natural eyes. As a result, it is possible to appropriately correct red eyes generated in various modes.

  Note that by using the saturation, lightness, and hue as feature quantities, it is possible to easily set a function for obtaining a correction curve based on human sensitivity, and to obtain a more natural corrected image. it can.

  In addition, since the smoothing according to the feature area to be processed is performed after (or before) the correction of the feature area, the entire red-eye area can be appropriately smoothed. Furthermore, since correction is performed after increasing the resolution for a small feature region, it is possible to appropriately correct even for a small feature region.

<Modification>
Although one embodiment of the present invention has been described above, the present invention is not limited to the above embodiment, and various modifications can be made.

  In the above embodiment, correction is performed for each feature region. For example, when two adjacent iris regions are extracted, these iris regions may be treated as one feature region. The same applies to other types of feature regions.

  Further, although it has been described that the parameters are derived from the histogram, the concept of a histogram indicating the relationship between the feature value and the frequency may be expressed in an arbitrary manner in the program. For example, a histogram can be expressed as a plurality of variables, arrays, or objects.

  Although the values YH1, YH2, and YH3 have been described as being predetermined for the hue, these values may be obtained using a function. For example, the values YH1, YH2, and YH3 may be obtained by a function having the hue range and mode value before correction, the mode value after correction, and the like as parameters.

  In the above-described embodiment, the example in which the maximum value, the mode value, and the variance of the feature amount of interest are used as the parameters used for generating the correction curve and the smoothing filter, and the minimum value is used supplementarily has been described. Other parameters may be used. For example, an average value may be used instead of the mode value, and a median value (median) may be used. As described above, by using the statistical value derived from the histogram, it is possible to easily prepare parameters necessary for generating the correction curve and the smoothing filter.

In the above embodiment, hue, saturation, and lightness are used as feature amounts. However, values in a color system (color space) such as L ^* a ^* b ^* , LUV, and XYZ may be used as feature amounts. Of course, RGB values may be used as feature values as they are. Note that correcting the feature amount ultimately corresponds to correcting the pixel value.

  For smoothing, another method different from the above embodiment may be used. For example, the filter may be 5 × 5 or more, and the weighted average may be performed only on the adjacent unit regions in the four directions.

  On the other hand, the acquisition of image data in the image processing apparatus 1 is not read from the memory card 92 as in the above-described embodiment, but is performed between the image processing apparatus 1 and other apparatuses by, for example, cable connection, communication line, or radio. May be obtained by transmitting and receiving signals. The image processing device 1 may be provided with an imaging unit to acquire image data.

  The program 141 may be written in the ROM 15 in advance. In the image processing apparatus 1, a series of image processing is all executed by software processing by the CPU, but part or all of these processing can be realized by a dedicated circuit. In particular, rapid image processing is realized by executing repetitive operations in a logic circuit.

  In the above embodiment, the shape of the target region 402 is a rectangle, but is not limited thereto. For example, an elliptical shape or a shape arbitrarily designated by the user may be used. Furthermore, in the above embodiment, the target area 402 is specified by the user, but the target area 402 may be automatically determined by image recognition or determination based on a feature amount.

It is a figure which shows the external appearance of an image processing apparatus. It is a block diagram which shows the structure of an image processing apparatus. It is a block diagram which shows the function structure of an image processing apparatus. It is a block diagram which shows the function structure of a correction | amendment part. It is a figure which shows the flow of operation | movement of an image processing apparatus. It is a figure which shows the flow of operation | movement of an image processing apparatus. It is a figure which shows the flow of operation | movement of an image processing apparatus. It is a figure which illustrates the image to which the object area was specified. It is a figure which illustrates the characteristic area. It is a figure which shows the correction curve regarding saturation. It is a figure which shows the correction curve regarding brightness. It is a figure which shows the correction curve regarding a hue. It is a figure which shows the weighting coefficient of the filter for smoothing.

Explanation of symbols

1 Image processing device 61-63 Correction curve 91 Recording disk 141 Program 501 Pupil area 502a Iris area 503 Non-iris area S15, S21-S23, S32-S35, S42, S43 Steps

Claims (6)

A program for correcting red-eye in an image, the computer executing the program,
Performing red-eye correction on a plurality of feature regions extracted as red-eye regions from an image based on feature amounts;
Before or after the step of performing the red-eye correction, performing a smoothing according to the feature region to be processed for each of the plurality of feature regions;
A program characterized by having executed. The program according to claim 1, wherein execution of the program by the computer is performed on the computer.
Generating a histogram of feature amounts for each of the plurality of feature regions;
Determining the degree of smoothing based on parameters derived from the histogram;
Is further executed. A program for correcting red-eye in an image, the computer executing the program,
Determining a correction target area in the image;
Comparing the size of the correction target area with a predetermined value;
Increasing the resolution of the correction target area when the size of the correction target area is smaller than the predetermined value;
Performing red-eye correction on the correction target region;
A program characterized by having executed.   A computer-readable recording medium, wherein the program according to claim 1 is recorded. A red-eye correction method for correcting red-eye in an image,
Performing red-eye correction on a plurality of feature regions extracted as red-eye regions from an image based on feature amounts;
Before or after the step of performing the red-eye correction, performing a smoothing according to the feature region to be processed for each of the plurality of feature regions;
A red-eye correction method comprising: A red-eye correction method for correcting red-eye in an image,
Determining a correction target area in the image;
Comparing the size of the correction target area with a predetermined value;
Increasing the resolution of the correction target area when the size of the correction target area is smaller than the predetermined value;
Performing red-eye correction on the correction target region;
A red-eye correction method comprising:

Images