JP4163651B2 - Red-eye correction work support apparatus and program - Google Patents

Description

  The present invention relates to an apparatus and a program for supporting an operator's work required in the process of locally correcting a photographic image.

  When a person or animal is photographed with a stroboscope at night or in a dark place, the pupil (or part of the pupil) may be photographed in red or gold. For this reason, various methods have been proposed in which a pupil that has been photographed in red or gold (hereinafter referred to as “red-eye” including gold) is corrected to the original pupil color by digital image processing. .

  For example, Patent Document 1 discloses a method and apparatus for automatically recognizing red eyes based on the color, position, and size of a pupil from a region designated by an operator. Also, Patent Document 2 discloses a method of calculating a predetermined feature amount for each pixel in a region designated by an operator and selecting a portion having the most pupil feature as a correction target. However, in recognition processing based only on the features of the pupil part, it is difficult to distinguish an object having a local redness, such as red illumination, from red eyes. For this reason, it is difficult to perform all the processing automatically without the operator's operation.

On the other hand, Patent Document 3 discloses a method for detecting red eyes by a combination with face detection processing. In this method, red eyes can be automatically detected if the face can be accurately detected. However, it is difficult to detect red eyes without operator intervention in an image in which face detection is difficult, such as when the face is covered with a profile, hands or hair.
JP 2000-13680 A JP 2001-148780 A JP 2000-125320 A

  In view of the above circumstances, an object of the present invention is to provide a red-eye correction work support device and a program for reducing an operator's work load required in the process of detecting and correcting red eyes.

  In order to achieve the above-mentioned object, the present invention provides a red-eye correction work support device that supports the work of correcting red eye color by detecting red eyes in which at least a part of the pupil is displayed in red from an image. This red-eye correction work support apparatus obtains an image in which red-eye color is corrected, and a red-eye automatic detection unit that automatically detects red-eye, a confidence level calculation unit that calculates the confidence level of the detection result of red eyes by the red-eye automatic detection unit A process selection execution unit that selects and executes one process based on the confidence level calculated by the confidence level calculation unit from among a plurality of processes that are different from each other and the content of the operation requested by the user The process selection execution unit selects a process with a smaller operation burden on the user as the calculated degree of confidence is higher.

  Here, the degree of confidence is an index representing how reliable the judgment made in the process of red-eye detection is. For example, the determination process by a normal computer is performed by comparing the value to be determined with a predetermined threshold value. If the difference from the threshold value is large, the degree of confidence in the determination is strong, and the difference from the threshold value is large. If is small, it can be considered that the degree of confidence in the determination is weak.

  The process selection execution unit has a function of selecting and executing a process for requesting the user to confirm the detection result or the corrected image when the confidence level calculated by the confidence level calculation unit is lower than a predetermined threshold, for example. provide. As a process for requesting the confirmation operation, a process for outputting a predetermined sound can be considered.

  The process selection execution unit selects a process for registering images in a predetermined list when the degree of confidence calculated by the confidence level calculation unit is lower than a predetermined threshold, and the user performs confirmation operations collectively later. You may be able to do it.

  Further, the red-eye correction work support program of the present invention detects a red eye in which at least a part of the pupil is displayed in red from the image, and causes the computer to execute processing for supporting the work of correcting the color of the red eye. A work support program, comprising: a red eye automatic detection unit that automatically detects red eyes; a confidence level calculation unit that calculates a confidence level of a red eye detection result by the red eye detection unit; and an image with a corrected red eye color Function as a process selection execution unit that selects and executes one process based on the confidence level calculated by the confidence level calculation unit from among a plurality of processes that are different to the user and require different operations. The process selection execution unit is made to function so as to select a process with less user operation burden as the degree of confidence increases.

  According to the red-eye correction work support apparatus and program of the present invention, a case where a user's confirmation operation is necessary and a case where automatic processing is sufficient are automatically distinguished based on the degree of confidence. For this reason, the work burden on the operator is greatly reduced.

  In particular, in cases where the degree of confidence is low and a confirmation operation is required, if a process that alerts the user, such as voice output, is executed, the user can simply perform the requested operation without checking whether confirmation is necessary. Necessary confirmation can be performed only by performing.

  In addition, if a function for registering a case with a low confidence level that requires a confirmation operation is provided, it is possible to collectively confirm information stored in the list, which further improves usability.

  The function of the red-eye correction work support device of the present invention is realized mainly by a software program. Therefore, the red-eye correction work support device is not particularly limited in appearance and size as long as it has a storage unit that can store a program and image data and a calculation unit that can execute the stored program.

  Specifically, for example, if a red-eye correction work support program is installed in a general-purpose computer having a CPU, a memory, a hard disk, and other various input / output interfaces, the general-purpose computer functions as a red-eye correction work support device.

  Also, in the case of a dedicated machine such as a digital photographic printer, a red-eye correction work support function can be added to the apparatus if it can be executed by incorporating a red-eye correction work support program.

  Alternatively, it may be provided as a memory / logic mixed type semiconductor device. When provided as a semiconductor device, if the device is mounted on a digital camera or the like and operated, the digital camera functions as a camera with a red-eye correction work support function.

  In general, it is preferable to realize this by adding a red-eye correction work support process to a series of processes executed by the red-eye detection program or the red-eye correction program. Therefore, in the following description, the red-eye correction work support device of the present invention is disclosed through the processing of various devices having a red-eye correction work support function.

(Embodiment 1)
The first embodiment is an example in which a red-eye correction work support function is incorporated in an image processing apparatus installed in a laboratory that provides a photo print service. FIG. 1 shows an outline of red-eye detection processing and red-eye correction processing executed by the image processing apparatus.

  As shown in the flowchart of FIG. 1, the image processing apparatus acquires an image to be subjected to red-eye detection and correction processing (S101). The image is acquired by reading data stored in a hard disk or various media. Subsequently, red-eye detection (S102) and red-eye correction (S103) are performed on the acquired image.

  Subsequently, the degree of confidence in red-eye detection and correction processing is calculated (S104). The degree of confidence is an index that indicates whether red-eye detection has been performed accurately or whether an image has been corrected appropriately.

  If it is determined in step S105 that the calculated confidence level is greater than a predetermined threshold, the image corrected in step S103 is output as it is in step S109.

  If it is determined in step S105 that the calculated confidence level is equal to or less than a predetermined threshold, the image processing apparatus generates a notification sound and displays the image corrected in step S103 on the display screen of the image processing apparatus. (S106). When an input for instructing correction from the operator is detected (S107), the image is corrected in accordance with the instruction (S108), and then the corrected image is output (S109).

  On the other hand, if there is no input from the operator, the image corrected in step S103 is registered in the confirmation wait list together with the image before correction and the detection result obtained in step S102 (S110). When the above processing is completed, the process returns to step S101 to acquire the next image and repeat the same processing.

  FIG. 2 is a flowchart showing an outline of confirmation processing for an image registered in the confirmation waiting list. As shown in the figure, the image processing apparatus first displays a waiting list for confirmation on the screen (S201). When the operator selects one image from the list, the selection is accepted (S202), and the selected image is displayed on the display screen (S203). When an input for instructing correction from the operator is detected (S204), the image is corrected in accordance with the instruction (S205), and then the corrected image is output (S206). If there is no instruction input from the operator, the corrected image obtained in step S103 of FIG. 1 and stored in step S110 is output as it is.

  As is clear from the above description, in the above processing, when the degree of confidence in red-eye detection / correction processing is high, a corrected image is automatically output without intervention by the operator. On the other hand, when the degree of confidence is low, a notification sound for prompting the operator's intervention is generated, and correction from the operator is accepted. In addition, in preparation for the case where the operator cannot immediately perform the correction work, a function is also provided in which the operator once registers in the confirmation waiting list so that the correction work can be performed later, and the confirmation is performed later.

  Therefore, the operator does not need to determine whether or not intervention is necessary while checking all the images one by one, and only has to pay attention to the images when notified. Furthermore, since it can be registered in the list and processed later, it is possible to proceed with work according to the convenience of the user. This greatly reduces the operator's burden.

  Hereinafter, the main steps constituting the flowchart of FIG. 1 will be described in more detail. First, the red-eye detection process in step S102 will be described. In the present embodiment, red eyes are detected by the method described below. However, the method of automatically detecting red eyes in the present invention is not particularly limited, and any known method can be applied.

  FIG. 3 is a flowchart showing an outline of the red-eye detection process. First, an image according to resolution is acquired (S301). FIG. 4 is a diagram for explaining resolution-specific images. In the present embodiment, as shown in the figure, an image 15 having the same resolution as the input image 10, an image 16 having a resolution ½ of the input image 10, and an image 17 having a resolution ¼ of the input image 10 are obtained in advance. Generated and stored in memory.

  An image 15 having the same resolution as the input image 10 is generated by copying the input image 10, and images 16 and 17 having a resolution different from that of the input image 10 are so-called thinning that reduces the number of pixels with respect to the data of the input image 10. Generated by executing the process. In step S301 in FIG. 3, the image by resolution is acquired by reading out the image by resolution stored in the memory.

Next, color space conversion processing is performed on the acquired resolution-specific image (S302). Specifically, the R (red), G (green), and B (blue) values of each pixel constituting the resolution-specific image are converted into Y (luminance) and Cb (green and blue color differences) according to a predetermined conversion formula. ), Cr (color difference between green and red), and Cr ^* (color difference between skin color and red) to convert the color system of the image.

Y, Cb, and Cr are coordinate systems generally used in JPEG images and the like, and Cr ^* is a coordinate axis that represents the direction in which red and skin color are best separated in RGB space. The direction of the coordinate axes is determined in advance by applying a linear discriminant analysis method to the red sample and the skin color sample. By defining such coordinate axes, the red-eye detection accuracy can be increased as compared with the case where detection is performed in a normal YCbCr space.

  Subsequently, a determination target range is set on the image after color space conversion (S303), and red-eye identification determination processing is performed for the determination target range (S304). FIG. 4 shows an overview of the target range setting process in step S303 and the identification determination process in step S304.

  FIG. 5A shows a state in which the determination target range 18 is set on the resolution-specific image 15 that has been color space converted in step S302. In the present embodiment, the determination target range 18 is an area of 13 pixels × 13 pixels, but for the sake of convenience of explanation, the determination target range 18 is shown larger in the figure.

  In the identification determination process, an image included in the set determination target range 18 is applied to a plurality of red-eye detectors, and an image included in the determination target range 18 is determined as a red eye from a combination of detection results of the plurality of detectors. Judge whether it can be estimated. If it can be estimated as red-eye, the area is detected as red-eye.

  However, here, the red-eye detector is an identification that outputs a parameter for calculating a feature quantity effective for distinguishing between red-eye and non-red-eye, and an identification point that represents the degree of red-eyeness by using the calculated feature quantity as input. Means a combination of threshold values determined so as to maintain a predetermined positive detection rate by calculating a cumulative addition value of the identification points by applying the detector and its parameters and the identifier to a large number of red-eye samples.

  Parameters and classifiers are determined by performing learning in advance using a large number of red-eye samples and non-red-eye samples. Learning can be performed using a known method such as a neural network method or a boosting method known as a machine learning technique.

  The samples used for learning are, for example, as shown in FIGS. 6A, 6B, and 6C, a sample in which the entire pupil is red, a sample in which 80% of the pupil is red, and a sample in which 60% of the pupil is red. Thus, it is preferable to give a predetermined variation to the size of the red region to be tightened in the unit rectangle.

  In addition, if the learning sample includes a sample in which the center of the red region is slightly shifted from the center of the unit rectangle, the red region can be extracted even in a shifted state. For this reason, it is possible to maintain the extraction accuracy even if the interval is set when scanning is performed by setting a determination target range for calculating the feature amount on the image, and the processing time can be shortened.

  Further, the threshold value can be correctly detected with a probability equal to or higher than a predetermined probability by applying the feature amount calculation parameter and the discriminator determined by learning to as many red-eye samples as possible to calculate the cumulative addition value of the discrimination points. It is good to define as follows.

  FIG. 7 is a flowchart showing specific processing steps of the identification determination processing in step S304 of FIG. The symbol i in the flowchart is a symbol for distinguishing red-eye detectors. When there are N red-eye detectors, 0 ≦ i ≦ N−1. N red-eye detectors, that is, a parameter i, a discriminator i, and a threshold value i (0 ≦ i ≦ N−1) used for calculating a feature value are stored in a memory or a hard disk.

  First, both i and the value of the cumulative identification point are initialized to 0 (S401). Next, using the feature quantity calculation parameter i, a feature quantity is calculated for the above-described determination target range 18, and a calculation result is obtained (S402). Next, an identification point is obtained by referring to the classifier i based on the calculation result (S403), and the identification point is added to the accumulated identification point (S404). Next, the accumulated identification point is compared with a threshold value i (S405). At this time, if the cumulative identification point is below the threshold value i, it is determined as non-red-eye.

  On the other hand, if the cumulative identification point exceeds the threshold value i, it is determined whether or not the processing has been completed for all the classifiers by determining whether i is N−1 (S406). If i is smaller than N-1, i is incremented by 1 (S407), and the processing from step S402 to S407 is repeated in the same manner. When the processing is completed for all the classifiers (S406), the image within the determination target range 18 is determined as a red-eye candidate region at that time, and is registered in the candidate list.

In this embodiment, the characteristic amount calculating parameters, reference channels (Y, Cb, Cr, ^Cr *), the feature quantity of type (pixel value itself, 2-point difference, 4-point difference), referenced pixel Consists of coordinates within the determination target range.

  The identification determination process described above is repeatedly executed while gradually shifting (scanning) the determination target range 18 as indicated by an arrow in the image 15 of FIG. The setting of the determination target range and the identification determination process are ended when it is determined that the scanning is ended in step S305 in the flowchart of FIG.

  Next, in step S306 of FIG. 3, it is determined whether or not the processing for all resolution images has been completed. If another image by resolution is unprocessed, the process returns to step S301, the next image 16 by resolution in FIG. 4 is acquired, and the same processing is repeated.

  The reason why the processing is repeated with different resolutions is as follows. FIG. 5B shows a state in which the determination target range 18 is set on the image 16 having a lower resolution than the resolution-specific image 15. The size of the determination target range 18 is 13 pixels × 13 pixels as described above. When the resolution is lowered, the determination target range 18 includes a wider range than when the resolution is high.

  Therefore, for example, as shown in FIGS. 5A and 5B, when the image of the pupil 19 is included in the image, the pupil that could not be detected by the identification processing for the resolution-specific image 15 in FIG. 19 may be detected by the identification processing for the resolution-specific image 16. Information on the resolution of the image when the red eye can be detected is stored in a memory or the like.

  If it is determined in step S306 that the processing for all resolution-specific images has been completed, candidate organization processing is executed (S307). FIG. 8 is a diagram for explaining the candidate organization process. As shown in the figure, in the determination target range setting process and the identification determination process, one red eye may be detected as two red eyes.

  For example, when the red eye is the ellipse area 20 shown in FIG. 8, the identification determination performed for the determination target range 18b when the area 21a is determined to be red eye by the identification determination processing performed for the determination target range 18a. This is a case where the region 21b is determined to be red-eye in the processing. In such a case, the candidate rearrangement process is a process of leaving only the region with the higher identification point among the red-eye 21a and the red-eye 21b and deleting the other from the list.

  The center coordinates and size of the region finally left by the above processing are output as a red-eye list.

  Subsequently, the red-eye correction process in step S103 of FIG. 1 will be described. In the present embodiment, for each red eye detected by the red eye detection process, a pixel having a color difference Cr value exceeding a predetermined value is extracted, and the area is shaped by the morphology process to form the shaped area. The color of each pixel is replaced with a color suitable for the pupil, such as gray having a predetermined brightness. However, in the present invention, the red-eye automatic correction method is not particularly limited, and any known method can be applied.

  Next, the confidence level calculation process in step S104 will be described. In the present embodiment, the degree of confidence is calculated using an identification point calculated in the course of the red-eye detection process. In the red-eye detection process, as described above, if the identification point is greater than or equal to the threshold value i, it is determined as red-eye, and if it is below the threshold value i, it is determined as non-red-eye. However, in this method, the difference between the identification point and the threshold value is small. It is considered that the reliability of the determination result is higher when the difference between the identification point and the threshold value is larger than when it is. Therefore, if the confidence level is defined in advance to be high when the difference is large and low when the difference is small, the confidence level can be obtained as an index representing the reliability of the detection after the detection process. it can.

  Next, the determination process in step S105 will be described. In step S105, as described above, it is determined whether or not the calculated confidence level is greater than a predetermined threshold value. Although the threshold value may be a constant value at all times, the effect of false detection / correction changes depending on the application of the image.Therefore, in this embodiment, determination is performed after setting an optimal threshold value for each application. Is going.

  For example, when a red-eye corrected image is printed out, erroneous detection and correction are more conspicuous as the print size increases. Therefore, in the present embodiment, the threshold value is set so as to increase as the print size increases.

  In addition, even if the print size is not large, even if the proportion of red eyes in the image is large, such as a photograph of a person's face that is up, misdetection and correction are conspicuous in the same way. It is set.

  Furthermore, in the case of a service for digitizing an image recorded on a film and storing it on a medium or the like, the image may be used for any purpose in the future. Therefore, in such a service, determination is performed by setting a high threshold value so that the influence of erroneous detection or correction is not a concern regardless of the use of the image.

  The process of step S106 is a process for generating a notification sound and displaying a corrected image on the screen in the present embodiment, but is another process as long as it can call the operator's attention. Also good. For example, a method of calling attention by blinking the entire screen instead of sound can be considered. Further, the notification method may be changed step by step according to the degree of confidence. For example, the volume may be changed according to the degree of confidence, or the time for displaying a screen and waiting for a correction instruction input may be changed according to the degree of confidence.

  Next, the processing of steps S107 and S108 will be described. As the correction instruction input from the user, for example, the following instruction can be considered. One is an instruction to remove erroneous detection. In the present embodiment, when, for example, electrical decoration other than red-eye is detected as red-eye and has been corrected to the eye color, a desired area on the screen is designated so that the user can request correction. Menus, buttons, etc. for inputting means and correction cancellation instructions are displayed on the screen. Furthermore, the function which a user can correct directly the location corrected accidentally is also provided.

  On the other hand, red eyes leaking from detection are specified directly on the screen to instruct correction, or an area including red eyes leaking from detection is specified, and red-eye re-detection processing is executed in that limited area, automatically It also provides the ability to make corrections. Alternatively, the red-eye automatic detection and automatic correction processing may be re-executed without particularly specifying the region. If the detection process is performed with the threshold of the identification point set to a lower value at the time of re-execution, the number of red-eye detections can be increased and undetected.

  When outputting the corrected image in step S109, an image obtained by directly rewriting the image acquired in step S101 may be output, or difference information between the original image and the corrected image may be output. Also good. In the latter case, a correction image is formed by synthesizing the original image and difference information in an apparatus (such as a printer) that uses the result.

(Embodiment 2)
As in the first embodiment, the second embodiment is an example in which a red-eye correction work support function is incorporated in an image processing apparatus installed in a laboratory that provides a photo print service. However, in the present embodiment, the first image processing apparatus incorporating the red-eye correction work support function is different from the second image processing apparatus that executes the red-eye correction process.

  FIG. 9 shows an outline of the red-eye detection process executed by the first image processing apparatus. As shown in the flowchart, the image processing apparatus acquires an image to be subjected to red-eye detection processing (S501). The image is acquired by reading data stored in a hard disk or various media. Subsequently, a red-eye detection process is automatically performed on the acquired image (S502).

  Subsequently, the degree of confidence in red-eye detection is calculated (S503). If it is determined in step S504 that the calculated degree of confidence is greater than a predetermined threshold, the result detected in step S502 is output as it is in step S508.

  If it is determined in step S504 that the calculated degree of confidence is equal to or less than a predetermined threshold, the image processing apparatus generates a notification sound and detects the red eye such as the position and size of the red eye detected in step S502. The result is displayed on the display screen (S505). When an input for instructing correction from the operator is detected (S506), the detection result is corrected according to the instruction (S507), and then the detection result is output (S508).

  On the other hand, if there is no input from the operator, the red eye detection target image in step S502 is registered in the confirmation wait list together with the red eye detection result (S509). When the above process is completed, the process returns to step S501, the next image is acquired, and the same process is repeated.

  FIG. 10 is a flowchart showing an outline of confirmation processing for an image registered in the confirmation waiting list. As shown in the figure, the image processing apparatus first displays a confirmation waiting list on the screen (S601). When the operator selects one image from the list, the selection is accepted (S602), and the result of red-eye detection for the selected image is displayed on the display screen (S603). If an input for instructing correction from the operator is detected (S604), the image is corrected according to the instruction (S605), and then the corrected detection result is output (S606). If there is no instruction input from the operator, the detection result obtained in step S502 of FIG. 9 is output as it is.

  FIG. 11 is a flowchart showing an outline of red-eye correction processing executed by the second image processing apparatus. As shown in the figure, the second image processing apparatus acquires the image and red-eye detection result output by the first image processing apparatus (S701). Then, the detected red eye is corrected (S702), and a corrected image is output (S704).

  Note that the red-eye detection and red-eye correction method and the confidence level calculation method are the same as those in the first embodiment, and thus description thereof is omitted.

  The first embodiment is a mode in which the user confirms the image after being automatically corrected and inputs a correction instruction, but the second embodiment only includes a detection result before the image is corrected. If the confidence level is low at the stage, the user is prompted for confirmation. Since the operator corrects misrecognition and the like before executing the image correction process, the burden on the apparatus is less than that in the first embodiment.

(Embodiment 3)
The third embodiment is an example in which a red-eye correction work support function is incorporated in a digital camera. FIG. 12 shows an outline of red-eye detection processing and red-eye correction processing executed by the digital camera.

  As shown in the flowchart of FIG. 12, the digital camera obtains an image to be subjected to red-eye detection and correction processing by capturing or reading from the memory (S801). Subsequently, red-eye detection (S802) and red-eye correction (S803) are performed on the acquired image.

  Subsequently, the degree of confidence in red-eye detection and correction processing is calculated (S804). If it is determined in step S805 that the calculated degree of confidence is greater than a predetermined threshold, the corrected image is output to a liquid crystal monitor provided in the digital camera (S806).

  If it is determined in step S805 that the calculated confidence level is equal to or less than a predetermined threshold, the digital camera generates a notification sound and displays the image corrected in step S803 on the liquid crystal monitor. At this time, since the monitor size of the digital camera is small and difficult to check, when displaying a corrected image, it is easy to check by displaying a frame with a frame around the correction area or enlarging the correction area. (S807).

  When an input for instructing correction from the operator is detected (S808), the image is corrected according to the instruction (S809), and then the corrected image is output (S810).

  On the other hand, if there is no input from the operator, the image corrected in step S803 is stored as it is as a corrected image (S810). When the above processing is completed, the process returns to step S801 to acquire the next image and repeat the same processing.

  Note that the red-eye detection and red-eye correction method and the confidence level calculation method are the same as those in the first embodiment, and thus description thereof is omitted.

  In the case of a digital camera, particularly when the liquid crystal monitor is small, it is difficult to determine whether the image requires a correction operation. For this reason, if the image which needs correction and the image which does not need correction can be automatically distributed based on the confidence degree of a detection process like the apparatus of this Embodiment, a user's burden will be reduced.

Flow chart showing an example of red-eye detection / correction processing Flow chart showing an overview of the confirmation process for registered images Flow chart showing the outline of the red-eye detection process Illustration for explaining images by resolution The figure which shows the outline of judgment object range setting processing and identification judgment processing Diagram showing an example of a sample used for red-eye detector learning Flow chart showing specific processing steps of identification determination processing Diagram for explaining candidate organization process Flow chart showing another example of red-eye detection processing Flowchart showing an overview of processing for confirming registered detection results Flow chart showing the outline of red-eye correction processing Flow chart showing another example of red-eye detection / correction processing

Explanation of symbols

10 Input images 15, 16, 17 Resolution-specific images 18, 18a, 18b, 23 Determination target range 19 Pupil 20 Red-eye 21a, 21b Red-eye candidate

Claims (5)

A red-eye correction work support device that supports the work of detecting red eyes in which at least part of a pupil is displayed in red from an image and correcting the color of the red eyes,
A red-eye automatic detection unit that automatically detects red-eye,
A confidence level calculation unit that calculates the confidence level of the detection result of the red eye by the red eye automatic detection unit using a value obtained in the process of the automatic detection ,
A process for obtaining an image in which the color of the red eye is corrected, and one process based on the confidence level calculated by the confidence level calculation unit is selected from among a plurality of processes having different operation contents requested from the user. A process selection execution unit for selecting and executing,
The red eye correction work support apparatus, wherein the process selection execution unit selects a process with less operation burden on the user as the degree of confidence is higher.
  The process selection execution unit selects a process for requesting the user to confirm the detection result and / or the corrected image when the confidence level calculated by the confidence level calculation unit is lower than a predetermined threshold. The red-eye correction work support apparatus according to claim 1, wherein the red-eye correction work support apparatus is executed.   The red-eye correction work support apparatus according to claim 1, wherein the process of requesting the confirmation operation is a process of outputting a predetermined sound.   The processing selection execution unit selects processing for registering the image in a predetermined list when the confidence level calculated by the confidence level calculation unit is lower than a predetermined threshold. 4. The red-eye correction work support device according to any one of 3 above. A red-eye correction work support program for causing a computer to execute a process for supporting the work of detecting a red eye in which at least part of a pupil is displayed in red from an image and correcting the color of the red eye.
Red-eye automatic detection unit that automatically detects red-eye,
A confidence level calculation unit that calculates the confidence level of the detection result of the red eye by the red eye automatic detection unit by using a value obtained in the process of the automatic detection , and for obtaining an image in which the color of the red eye is corrected It functions as a process selection execution unit that selects and executes one process based on the confidence level calculated by the confidence level calculation unit from among a plurality of processes that differ in the content of the operation requested from the user. ,
A red-eye correction work support program that causes the process selection execution unit to function so as to select a process with less operational burden on the user as the degree of confidence increases.