JP2012243206A - Image processing method, image processor and image processing program - Google Patents

Abstract

PROBLEM TO BE SOLVED: To provide an image processor for achieving quick face detection.SOLUTION: The image processing method includes: a step of generating a smoothed image from an acquired image; a step of calculating the mean value of respective pixels on the basis of the images of a plurality of frames before a processing object image including the processing object image, and calculating a distribution value from the calculated mean value of the pixels and the pixel value of each frame; a step of executing further smoothing processing to the luminance value of the smoothed image of the processing object image to generate an image for normalization, and normalizing the distribution value image by using the value of each pixel of the generated image for normalization to generate a normalized distribution value image with respect to each pixel value of the generated distribution value image; a step of executing expansion processing to the normalized distribution value image to generate a corrected distribution value image; a step of executing arbitrary threshold processing to each pixel value of the corrected distribution value image to generate a binary image, and for detecting a figure area; and a step of executing face detection processing to the pixels whose distribution value is determined so as to be larger than a fixed value.

Description

  The present invention relates to an image processing device that detects a person's face from a time-series image taken by an imaging device, limits the detection range by detecting the person's movement, and detects the face at high speed. The present invention relates to a processing method and an image processing program.

  Conventionally, a method described in Non-Patent Document 1 is known as a method of detecting a face from an image or video. Further, a method described in Non-Patent Document 2 is known as a method for estimating the detected face posture. By using these methods, it is possible to measure the degree of attention to digital signage as an objective advertising effect index of digital signage that has become popular in recent years.

Yuji Mita, Toshimitsu Kaneko, Osamu Hori, “Probabilistic Incremental Code Correlation Suitable for Image Matching of Individual Differences” IEICE Transactions D-II Vol. J88-D-II No. 8 pp. 1614-1623, The Institute of Electronics, Information and Communication Engineers 2005 Shingo Ando, Yoshinori Kusachi, Akira Suzuki, Kenichi Arakawa, “Pose Estimation Method for 3D Objects Using Support Vector Regression”, IEICE Transactions D Vol. J89-D No. 8 pp. 1840-1847, The Institute of Electronics, Information and Communication Engineers 2006

  However, a general face detection method is a method for detecting a face from an entire image or video, and when using image processing technology that uses advanced feature amounts to suppress false detection and over-detection, the processing speed There is a problem that will decrease.

  The present invention has been made in view of such circumstances, and by limiting the detection range of a person's face, high-speed face detection can be realized while suppressing false detection and excessive detection. An object is to provide an image processing apparatus, an image processing method, and an image processing program.

  The present invention provides a smoothed image generation unit, a variance value image generation unit, a variance value image normalization unit, an expansion / contraction processing unit, a person area detection unit, a face, An image processing method in an image processing apparatus including a detection processing unit, wherein the smoothed image generation unit performs a smoothing process on the acquired image to generate a smoothed image, and the variance value The image generation unit calculates the average value of each pixel based on images of a plurality of frames before the processing target image including the processing target image, and distributes the calculated average value of each pixel and the corresponding pixel value of each frame. A variance value image generation step for calculating a value, and the variance value image normalization unit performs a further smoothing process on the luminance value of the smoothed image of the processing target image to generate a normalization image; Each pixel value of the generated variance value image On the other hand, a normalized image generation step of generating a normalized variance value image by normalizing the variance value image using the value of each pixel of the generated normalization image, and the expansion / contraction processing unit includes the normalization An expansion / contraction step for generating a corrected dispersion value image by performing an expansion process on the normalized dispersion value image, and the person region detection unit may arbitrarily set each pixel value of the correction dispersion value image. A human region detection step of detecting a human region by generating a binary image by performing threshold processing, and the face detection processing unit performs face detection processing on pixels for which the variance value is determined to be greater than a predetermined value. And a face detection step to be performed.

  The present invention includes a variance value table generation step of generating, as a variance value table, a variance value that is minimum for each luminance value of the processing target image with respect to the variance value of each pixel generated by the variance value image generation step. Furthermore, each pixel value in the person area detection step is determined based on the variance value table generated in the variance value table generation step and each pixel value of the normalization image generated in the normalized image generation step. A threshold value is automatically calculated and threshold processing is performed.

  The present invention further includes a face detection position holding step for holding a face detection position with respect to the image of the previous frame detected in the face detection step, and the face detection position held in the face detection position holding step. , The face detection is performed in the face detection step by giving priority to the direction in which the pixel value of the corrected dispersion value image generated in the expansion / contraction step is higher, and the variance around the corresponding pixel position in the person region detection step Even when the value is determined to be equal to or less than a certain value, face detection is performed around the corresponding pixel position.

  The present invention is an image processing device that performs face detection from an acquired image, and performs smoothing processing on the acquired image to generate a smoothed image, and the processing target image including the processing target image A variance value image generating means for calculating an average value of each pixel based on the images of a plurality of frames and calculating a variance value from the calculated average value of each pixel and a corresponding pixel value of each frame; Further smoothing processing is performed on the luminance value of the smoothed image of the image to generate a normalization image, and for each pixel value of the generated variance value image, each of the generated normalization image Normalized image generation means for generating a normalized dispersion value image by normalizing the dispersion value image using the pixel value, and correction dispersion by performing dilation processing on the normalized normalized dispersion value image Expansion and contraction means for generating a value image; Human area detection means for detecting a person area by performing arbitrary threshold processing on each pixel value of the corrected variance value image to generate a binary image, and pixels for which the variance value is determined to be greater than a predetermined value And a face detection means for performing face detection processing on the image.

  The present invention is an image processing program that causes a computer on an image processing apparatus that performs face detection from an acquired image to perform image processing, and performs smoothing processing on the acquired image to generate a smoothed image The average value of each pixel is calculated based on the image of a plurality of frames before the processing target image including the processing target image, and the variance value is calculated from the calculated average value of each pixel and the corresponding pixel value of each frame. A variance value image generation step to be calculated, and further smoothing processing is performed on the brightness value of the smoothed image of the processing target image to generate a normalization image, and each pixel value of the generated variance value image On the other hand, a normalized image generation step of generating a normalized variance value image by normalizing the variance value image using the value of each pixel of the generated normalization image, and the normalized normalized variance value image Against An expansion / contraction step for generating a corrected variance value image by performing a tension process, and an arbitrary threshold value processing for each pixel value of the corrected variance value image to generate a binary image, thereby detecting a human region It is characterized in that the computer is caused to perform a human region detection step and a face detection step of performing a face detection process on pixels for which the variance value is determined to be larger than a predetermined value.

  According to the present invention, by limiting the person area based on the person's motion information, it is possible to limit the area for estimating the face and posture of the person of the prior art and perform high-speed detection. Also, smoothing reduces noise in low-brightness areas, normalizes the variance value, which is the amount of motion based on the spatial brightness value of the entire image, and automates thresholds when detecting human areas by creating a table of variance values for each brightness value By performing the above and the like, it is possible to reduce excessive detection of a human area due to noise, omission of detection of a low luminance area, and the like. Furthermore, it becomes possible to perform high-speed search by prioritizing search areas according to the amount of movement of the person, and by detecting the face position from around the corresponding position based on the face position information detected in the previous frame An effect is obtained that the face of a stationary person with a small amount of motion can be detected.

It is a block diagram which shows the structure of one Embodiment of this invention. It is a flowchart which shows the processing operation of the apparatus shown in FIG. It is a flowchart which shows the processing operation of the apparatus shown in FIG. It is a flowchart which shows the processing operation of the apparatus shown in FIG. It is explanatory drawing which shows an example of a some frame and a dispersion value. It is explanatory drawing which shows an example of the expansion process by a morphological operation, and a degeneracy process.

  First, the principle of image processing will be described. Here, in order to limit the area before detecting a person's face, a person-like area is detected from an image or video. As the personness, a temporal change of each pixel component of the target image is used. To measure temporal changes, select multiple consecutive images including the image to be processed from the time-series images, and change the temporal changes for each pixel by calculating the variance of the same pixel in multiple frames. Calculate the amount.

  FIG. 5 shows a conceptual diagram of the change amount calculation. As shown in FIG. 5, the variance value is large in the region with movement, but the variance value is very small because the change in the pixel value is small in the region without movement. The calculation of the variance value uses a plurality of images including the image to be processed. Here, in the example of FIG. 5, the image to be processed is the Nth image, and the past M images based on the Nth image The variance value is calculated using the image up to the eye. At this time, when the number of images used for the variance calculation is small, the contribution ratio of one image increases, and the variance value greatly fluctuates with respect to an instantaneous change. For this reason, the variance value fluctuates greatly even in an area where there is originally little change.

  Therefore, a sufficiently large number of images are used for calculating the variance value. Here, when the luminance of the pixel is low due to the shooting environment or camera setting, there is a problem that the influence of noise during imaging is relatively greater than the change in the pixel value. Therefore, by performing spatial smoothing on the entire image, changes in pixel values due to the influence of noise are reduced. Also, the pixel value is small, such as when the brightness is spatially different within the same image due to shadows or backlighting, or when the reflectance of the object itself in the image is low and the pixel value in the image becomes small In the region, there is a problem that the amount of change measured is small. Therefore, by smoothing the brightness of all the pixel values with a brightness image obtained by spatially smoothing the brightness value of the image, and calculating the variance value, even in a region with a small pixel value, Measure changes.

  When the person stays at the same place for a long time, the person peripheral area has a movement due to the wobbling of the person, so the change of the pixel value becomes large and the variance value also becomes large. However, in a region near the center of the person where the variation of the pixel value is spatially small, the variance value is also small. Therefore, expansion and contraction processing of the morphology operation shown in FIG. 6 is performed on the calculated dispersion value, and hole filling processing near the center of the person where the dispersion value becomes small is performed. By performing the dilation process, the same large variance value as the person peripheral area can be obtained from the person peripheral pixel to the person center or the area outside the person. Only the variance value can be a smaller outer variance value. In this way, a variance value image having a large variance value in the person area and a small variance value in the other areas is generated.

  In order to extract a person region using the pixel value of the variance value image as a person, it is necessary to perform threshold processing based on the person-like value. Here, as described above, in order to further reduce the influence of noise or the like included in the imaging device, the minimum value of the variance value is tabulated for each luminance value of the pixel at the time of measurement, and the minimum variance value that is tabulated is determined according to the table. By calculating the threshold value, the threshold value corresponding to the luminance value is dynamically set. In addition, since the change in the pixel value of the person area is larger than that of the area other than the person, the priority is given to the search area for detecting the face of the person depending on the value of each pixel of the variance value image. Do a faster search. Furthermore, since the amount of movement decreases when the person is stationary, the face around the corresponding position even when the amount of movement is extremely small based on the position information where the face was detected in the previous frame. By performing the detection processing, omissions in detection of human faces are reduced.

<First Embodiment>
Hereinafter, an image processing apparatus according to a first embodiment of the present invention will be described with reference to the drawings. FIG. 1 is a block diagram showing the configuration of the embodiment. In FIG. 1, reference numeral 1 denotes a camera as an image input device. Reference numeral 2 denotes a smoothed image generating unit for performing spatial smoothing on the image acquired by the camera 1. An image storage unit 3 stores the image generated by the smoothed image generation unit 2. Reference numeral 4 denotes a dispersion value image generation unit that calculates an average value for each pixel of the image smoothed by the smoothed image generation unit 2 and generates a dispersion value image. Reference numeral 5 denotes a luminance value smoothing process performed on the smoothed image generated by the smoothed image generating unit 2, and the variance of each pixel calculated by the variance value image generating unit 4 according to the smoothed luminance value It is a dispersion value image normalization part which normalizes a value.

  An expansion / contraction processing unit 6 performs expansion processing / contraction processing on the dispersion value normalized by the dispersion value image normalization unit 5. Reference numeral 7 denotes a person area detection unit that detects a person area by performing threshold processing on the normalized dispersion value that has been subjected to expansion processing and contraction processing by the expansion / contraction processing unit 6. A face detection processing unit 8 performs face detection processing from the person area detected by the person area detection unit 7. Each process in the smoothed image generation unit 2, the variance value image generation unit 4, the variance value image normalization unit 5, the expansion / contraction processing unit 6, the person area detection unit 7, and the face detection processing unit 8 is performed by, for example, a computer or It is executed by hardware inside the camera. The image storage unit 3 may use a recording medium such as a hard disk, a RAID device, a CD-ROM, or a remote data resource via a network.

  Next, the processing operation of the image processing apparatus shown in FIG. 1 will be described with reference to FIG. FIG. 2 is a flowchart showing the processing operation of the image processing apparatus shown in FIG. 1 shown in FIG. First, the overall flow will be described. The smoothed image generating unit 2 smoothes the image captured by the camera 1 and stores it in the image storage unit 3 (step S1). Next, the variance value image generation unit 4 and the variance value image normalization unit 5 capture images stored in the image storage unit 2 (step S2). Then, the variance value image generation unit 4 calculates an average value for each pixel from the smoothed images of a plurality of frames (step S3), and calculates a variance value for each pixel (step S4). The variance value image generation unit 4 determines whether or not the processing has been completed for all the pixels (step S5). Generate a value image.

  Next, the variance value image normalization unit 5 generates a normalization luminance image in which the luminance value is smoothed from the acquired smoothed image of the processing target image (step S6), and the variance value image generation unit 4 The dispersion value image is normalized from the generated dispersion value image and the normalization luminance image generated in step S6 (step S7). Subsequently, the expansion / contraction processing unit 6 performs expansion processing and contraction processing on the normalized dispersion value image (step S8), and the person area detection unit 7 converts the dispersion value image after the expansion processing and contraction processing into the dispersion value image. The threshold value processing is performed for the person area (step S9), thereby detecting the person area and outputting the obtained person area image (step S10). The face detection processing unit 8 detects a person's face from the person area image (step S11).

Next, the operation of each process will be described in detail. An image acquired via the camera 1 is subjected to a smoothing process in the smoothed image generator 2. For the smoothing process, a filter process used by general image processing such as a Gaussian filter is used. The smoothed image is accumulated in the image accumulating unit 3, and the M value continuous image including the processing target image is acquired from the image accumulating unit 3 in the variance value image generating unit 4. Here, M is an arbitrarily large number of sheets. Next, for each pixel of the acquired M images, an average value ￣I (x, y) (￣ is attached to the head of I) of the pixel at the coordinate (x, y) according to the equation (1). calculate. Here, k is the 0th frame of the processing target image, the frame number up to the past M frame, and is the pixel value of the coordinates (x, y) of the frame number k of each frame.

After calculating the average value of each pixel, the variance value is calculated according to equation (2).

After performing the same processing for all the pixels, a processing target image is acquired from the image storage unit 3, and smoothing processing is performed on the luminance value of the processing target image. The smoothing process here uses a filter process such as a Gaussian filter as in the smoothed image generation unit. By dividing the value of the variance value image S (x, y) by the luminance value of the processing target image after the smoothing process, the variance value normalization process is performed according to equation (3). Here, it is assumed that the pixel value of the smoothed dispersion value image is G (x, y), and the dispersion value after the normalization process is S ′ (x, y).
S ′ (x, y) = S (x, y) / G (x, y) (3)

After performing the normalization process of the variance value image, the expansion process and the contraction process of the morphology operation are performed. In the dilation processing, for any pixel S ′ (x, y) in the normalized dispersion value image, an operation is performed in which the maximum value among the adjacent 8 pixels is the value of the pixel S ′ (x, y). Implement any number of times. Let S ″ (x, y) be the pixel value of the dispersion value image after the expansion processing. Similarly, in the contraction process, for any pixel S ″ (x, y) in the dispersion value image after the expansion process, the minimum pixel value among the adjacent eight pixels is set as the pixel S ″ (x , Y) is performed an arbitrary number of times. Let S ′ ″ (x, y) be the pixel value of the dispersion value image after the shrinkage processing. Here, the expansion processing and the contraction processing are executed the same number of times. Finally, the value of the variance value image S ′ ″ (x, y) is determined by an arbitrary threshold value in the equation (4), and face detection is performed only for pixels having a value exceeding the threshold value, thereby performing high-speed detection. Can be done. Here, f (x, y) is a pixel value of the binary image indicating whether or not the threshold value is exceeded. When f (x, y) = 1, the threshold value is exceeded, and f (x, y). When = 0, it indicates that the threshold is not exceeded. TH is an arbitrary threshold value.

<Second Embodiment>
Next, the processing operation of the image processing apparatus according to the second embodiment will be described with reference to FIG. FIG. 3 is a flowchart illustrating the processing operation of the image processing apparatus according to the second embodiment. In FIG. 3, the same processing operations as those shown in FIG.

  In the first embodiment, an arbitrary threshold value is set for the dispersion value image after the contraction process, and when the threshold value is exceeded, the corresponding pixel is set to 1 and the other pixel values are set to 0. doing. In the processing operation shown in FIG. 3, a table having the minimum variance value is prepared according to the luminance value at the corresponding position of the processing target image after the smoothing process, and after calculating the variance value of each pixel of the processing target image, A processing operation (step S12) is performed in which the luminance value of the pixel is compared with the value of the corresponding luminance value in the table, compared with the calculated variance value, and if the variance value is smaller, the table value is updated.

  Further, when the threshold processing (step S9) is performed, the threshold value of the corresponding pixel is determined from the table value corresponding to the luminance value of the pixel at the corresponding position of the processing target image and the corresponding pixel value of the normalization luminance image. By performing the automatic generation, threshold processing is performed in consideration of pixel value variation due to noise for each luminance value of the imaging device such as a camera. In the case of a general camera, noise characteristics differ from camera to camera, and it is difficult to exclude the influence of the camera noise on the dispersion value by an arbitrary threshold value. Here, since the variance value changes relatively depending on the size of the luminance value, and the variance value tends to increase in the person area, by holding a minimum variance value depending on the position and the luminance as a table, Even when the corresponding position is an area other than a person, it is possible to eliminate the minimum variance value due to noise and perform optimum threshold processing.

<Third Embodiment>
Next, the processing operation of the image processing apparatus according to the third embodiment will be described with reference to FIG. FIG. 4 is a flowchart showing the processing operation of the image processing apparatus according to the third embodiment. In FIG. 4, the same processing operations as those shown in FIG.

  In the first and second embodiments, face detection processing is performed for pixels whose pixel value of the person area image is 1. In the processing operation shown in FIG. 4, a processing operation for storing the position information of the face of the person detected in the previous frame in the buffer as a face detection result (step S13) is added, and the dispersion value image after the expansion / contraction processing is added. A search is made for the vicinity of the face position of the person detected in the previous frame from the pixel value, and the search range is narrowed down by searching for a face image preferentially from the direction of larger variance value, thereby performing faster face detection. Furthermore, even when the movement of the person is very small, the pixel value of the person area image around the detected face position is 0 based on the position information of the face detected in the previous frame. The detection of the human face is reduced by setting the corresponding position as the search target.

  As described above, by limiting the person area based on the person's motion information, it is possible to limit the area for estimating the face and posture of the person in the prior art and perform high-speed detection. Also, smoothing reduces noise in low-brightness areas, normalizes the variance value, which is the amount of motion based on the spatial brightness value of the entire image, and automates thresholds when detecting human areas by creating a table of variance values for each brightness value By performing the above and the like, it is possible to reduce excessive detection of a human area due to noise, omission of detection of a low luminance area, and the like. Furthermore, it becomes possible to perform high-speed search by prioritizing search areas according to the amount of movement of the person, and by detecting the face position from around the corresponding position based on the face position information detected in the previous frame It is also possible to detect the face of a stationary person with a small amount of movement.

  Note that a program for realizing the function of the processing unit in FIG. 1 is recorded on a computer-readable recording medium, and the program recorded on the recording medium is read into a computer system and executed to perform image processing. May be. Here, the “computer system” includes an OS and hardware such as peripheral devices. The “computer-readable recording medium” refers to a storage device such as a flexible medium, a magneto-optical disk, a portable medium such as a ROM and a CD-ROM, and a hard disk incorporated in a computer system. Further, the “computer-readable recording medium” refers to a volatile memory (RAM) in a computer system that becomes a server or a client when a program is transmitted via a network such as the Internet or a communication line such as a telephone line. In addition, those holding programs for a certain period of time are also included.

  The program may be transmitted from a computer system storing the program in a storage device or the like to another computer system via a transmission medium or by a transmission wave in the transmission medium. Here, the “transmission medium” for transmitting the program refers to a medium having a function of transmitting information, such as a network (communication network) such as the Internet or a communication line (communication line) such as a telephone line. The program may be for realizing a part of the functions described above. Furthermore, what can implement | achieve the function mentioned above in combination with the program already recorded on the computer system, what is called a difference file (difference program) may be sufficient.

  When detecting a person's face from a time-series image taken by an imaging device, the detection range can be limited by detecting the person's movement, and the detection can be applied to an indispensable use.

  DESCRIPTION OF SYMBOLS 1 ... Camera, 2 ... Smoothed image generation part, 3 ... Image storage part, 4 ... Dispersion value calculation image generation part, 5 ... Dispersion value image normalization part, 6 ... Expansion / contraction processing unit, 7 ... person area detection unit, 8 ... face detection processing unit

Claims (5)

In order to perform face detection from the acquired image, a smoothed image generation unit, a variance value image generation unit, a variance value image normalization unit, an expansion / contraction processing unit, a person region detection unit, a face detection processing unit, An image processing method in an image processing apparatus comprising:
A smoothing step in which the smoothed image generating unit performs a smoothing process on the acquired image to generate a smoothed image;
The variance value image generation unit calculates an average value of each pixel based on images of a plurality of frames before the processing target image including the processing target image, and calculates the average value of each pixel and the corresponding pixel of each frame. A variance value image generation step for calculating a variance value from the value;
The variance value image normalization unit performs a further smoothing process on the brightness value of the smoothed image of the processing target image to generate a normalization image, and sets each pixel value of the generated variance value image. On the other hand, a normalized image generation step of generating a normalized variance value image by normalizing the variance value image using the value of each pixel of the generated normalization image,
An expansion / contraction step in which the expansion / contraction processing unit generates a corrected dispersion value image by performing expansion processing on the normalized dispersion value image normalized; and
A human region detection step in which the human region detection unit detects a human region by performing arbitrary threshold processing on each pixel value of the corrected dispersion value image to generate a binary image;
An image processing method comprising: a face detection step in which the face detection processing unit performs a face detection process on a pixel for which the variance value is determined to be greater than a predetermined value.   A dispersion value table generating step for generating a dispersion value that is minimized for each luminance value of the processing target image as a dispersion value table with respect to the dispersion value of each pixel generated by the dispersion value image generation step; The threshold value of each pixel is automatically calculated in the person area detection step from the dispersion value table generated in the variance value table generation step and each pixel value of the normalization image generated in the normalized image generation step. The image processing method according to claim 1, wherein threshold processing is performed.   A face detection position holding step for holding a face detection position with respect to the image of the previous frame detected in the face detection step, with the face detection position held in the face detection position holding step as a reference; Priority is given to the direction in which the pixel value of the corrected variance value image generated in the expansion / contraction step is higher, face detection is performed in the face detection step, and the variance value around the pixel position in the person region detection step is a constant value. 3. The image processing method according to claim 1, wherein face detection is performed around the corresponding pixel position even when it is determined as follows. An image processing apparatus that performs face detection from an acquired image,
Smoothing means for smoothing the acquired image to generate a smoothed image;
A variance that calculates an average value of each pixel based on images of a plurality of frames before the processing target image including the processing target image, and calculates a variance value from the calculated average value of each pixel and the corresponding pixel value of each frame. A value image generating means;
A normalization image is generated by performing further smoothing processing on the luminance value of the smoothed image of the processing target image, and the generated normalization image is generated for each pixel value of the generated variance value image. Normalized image generation means for generating a normalized dispersion value image by normalizing the dispersion value image using the value of each pixel of the image;
Expansion and contraction means for generating a corrected dispersion value image by performing expansion processing on the normalized dispersion value image that has been normalized;
Human area detection means for detecting a human area by performing arbitrary threshold processing on each pixel value of the corrected dispersion value image to generate a binary image;
An image processing apparatus comprising: a face detection unit that performs face detection processing on a pixel for which the variance value is determined to be greater than a predetermined value. An image processing program for causing a computer on an image processing apparatus that performs face detection from an acquired image to perform image processing,
A smoothing step of performing a smoothing process on the acquired image to generate a smoothed image;
A variance that calculates an average value of each pixel based on images of a plurality of frames before the processing target image including the processing target image, and calculates a variance value from the calculated average value of each pixel and the corresponding pixel value of each frame. A value image generation step;
Further smoothing processing is performed on the brightness value of the smoothed image of the processing target image to generate a normalization image, and the generated normalization image is generated for each pixel value of the generated variance value image. A normalized image generation step of generating a normalized dispersion value image by normalizing the dispersion value image using a value of each pixel of the image;
An expansion / contraction step for generating a corrected dispersion value image by performing an expansion process on the normalized dispersion value image that has been normalized;
A person area detection step of detecting a person area by performing arbitrary threshold processing on each pixel value of the corrected dispersion value image to generate a binary image;
An image processing program that causes the computer to perform a face detection step of performing face detection processing on a pixel for which the variance value is determined to be greater than a predetermined value.

Images