JP2020198476A - Image processing device, program, and image processing method - Google Patents

Abstract

To keep the quality of an image in a boundary region between a portion including a large volume of information and a portion including a small volume of information of the image.SOLUTION: An image processing device according to one embodiment of the disclosed technique comprises a first division processing unit for dividing a region except the face region of a person included in an input image on the basis of a first feature amount of the input image; and an information volume reduction processing unit for executing a process of reducing an information volume, for each region obtained by the division by the first division processing unit.SELECTED DRAWING: Figure 8

Description

The present invention relates to an image processing apparatus, a program, and an image processing method.

With the spread of networks and communication devices, highly convenient communication using images such as moving images has become possible. In communication using such images, the quality of images and sounds may deteriorate when the network band is narrow. On the other hand, a method of detecting the congestion status of the network and reducing the amount of information of the image to be transmitted is used.

In addition, in an image processing device having a communication function using images such as a smartphone, a tablet terminal, and a video conferencing terminal, in order to reduce the amount of information to be communicated, a face or the like including the movement of the lips accompanied by talking in the image. There is disclosed a part that distinguishes between a highly important part and a less important part such as a background other than a human face and makes the amount of information in the image different (see, for example, Patent Document 1).

However, in the conventional apparatus, the quality of the image may deteriorate in the boundary region between the portion where the amount of information in the image is large and the portion where the amount of information in the image is small.

The disclosed technique has an object of maintaining the quality of an image in a boundary region between a portion having a large amount of information and a portion having a small amount of information in the image.

The image processing apparatus according to one aspect of the disclosed technology includes a first division processing unit that divides an area other than the face area of a person included in the input image based on the first feature amount of the input image, and the first division processing unit. Each area divided by the division processing unit has an information amount reduction processing unit that executes a process of reducing the information amount.

According to the disclosed technique, it is possible to maintain the quality of the image in the boundary region between the portion having a large amount of information and the portion having a small amount of information in the image.

It is an external view which shows an example of the structure of the video conferencing terminal which concerns on embodiment. It is a block diagram explaining an example of the hardware configuration of the video conferencing terminal which concerns on embodiment. It is a block diagram explaining an example of the functional structure of the video conferencing terminal which concerns on embodiment. It is a figure explaining the information amount reduction process by the video conferencing terminal which concerns on embodiment. It is a timing chart explaining the update timing for each area of the image in the video conferencing terminal which concerns on embodiment. It is a figure explaining the effect of reducing the information amount of an image by the video conferencing terminal which concerns on embodiment. It is a flowchart which shows an example of the process by the video conferencing terminal which concerns on embodiment. It is a block diagram explaining an example of the functional structure of the video conferencing terminal which concerns on 1st Embodiment. It is a figure explaining an example of the frame-shaped area set to surround a non-face area. It is a figure explaining an example of the division process of a non-face area. It is a figure explaining an example of the spatial filter used in the video conferencing terminal which concerns on 1st Embodiment, (a) is the figure explaining the spatial filter of size 3 × 3 pixels, (b) is the figure which explains the size 5 × 5. It is a figure explaining the spatial filter of a pixel. It is a flowchart which shows an example of the processing by the video conference terminal which concerns on 1st Embodiment. It is a block diagram explaining an example of the functional structure of the video conferencing terminal which concerns on 2nd Embodiment. It is a figure explaining an example of the division process of a frame-shaped area. It is a flowchart which shows an example of the processing by the video conference terminal which concerns on 2nd Embodiment.

Hereinafter, modes for carrying out the invention will be described with reference to the drawings. In each drawing, the same components may be designated by the same reference numerals and duplicate description may be omitted.

In the embodiment, a video conferencing terminal as an example of the image processing device will be described. However, the image processing device is not limited to the video conferencing terminal as long as it is a device having a communication function. The image processing device includes, for example, a PJ (Projector: projector), an IWB (Interactive White Board: a white board having an electronic whiteboard function capable of intercommunication), an output device such as a digital signage, a HUD (Head Up Display) device, and the like. Industrial machines, imaging devices, sound collectors, medical devices, network home appliances, automobiles (Connected Cars), notebook PCs (Personal Computers), mobile phones, smartphones, tablet terminals, game machines, PADs (Personal Digital Assistants), digital cameras, It may be a wearable PC, a desktop PC, or the like.

<Configuration of video conferencing terminal according to the embodiment>
First, an example of the configuration of the video conferencing terminal according to the embodiment will be described. FIG. 1 is an external view of the video conferencing terminal 10. As shown in FIG. 1, the video conferencing terminal 10 has a housing 1100, an arm 1200, and a camera housing 1300. Of these, the front side wall surface 1110 of the housing 1100 is provided with an intake surface formed by a plurality of intake holes, and the rear side wall surface 1120 of the housing 1100 is provided with an exhaust surface having a plurality of exhaust holes. 1121 is provided. As a result, the outside air behind the video conferencing terminal 10 can be taken in through the intake surface and exhausted to the rear of the video conferencing terminal 10 through the exhaust surface 1121 by driving the cooling fan built in the housing 1100. .. Since the sound collecting hole 1131 is formed on the right side wall surface 1130 of the housing 1100, the built-in microphone 114 can collect sounds such as voice, noise, and noise.

An operation panel 1150 is formed on the right side wall surface 1130 side of the housing 1100. The operation panel 1150 is provided with a plurality of operation buttons (108a to 108e), a power switch 109, and an alarm lamp 119, and a plurality of sound output holes for passing the output sound from the built-in speaker 115. The sound output surface 1151 formed by the above is formed. Further, on the left side wall surface 1140 side of the housing 1100, an accommodating portion 1160 is formed as a recess for accommodating the arm 1200 and the camera housing 1300. The right side wall surface 1130 of the housing 1100 is provided with a plurality of connection ports (1132a to 1132c) for electrically connecting the cable to the connection I / F 118. On the other hand, the left side wall surface 1140 of the housing 1100 is provided with a connection port for electrically connecting the cable 120c for the display 120 to the connection I / F 118.

Next, the arm 1200 is attached to the housing 1100 via the torque hinge 1210, and the arm 1200 can rotate in the vertical direction with respect to the housing 1100 within a tilt angle θ1 of 135 degrees. There is. FIG. 1 shows a state in which the tilt angle θ1 is 90 degrees. The camera housing 1300 is provided with a built-in camera 112, which can image users, documents, rooms, and the like. Further, a torque hinge 1310 is formed in the camera housing 1300. The camera housing 1300 is attached to the arm 1200 via a torque hinge 1310. The camera housing 1300 with respect to the arm 1200 in the vertical and horizontal directions within a pan angle θ2 of ± 180 degrees and a tilt angle θ3 of ± 45 degrees with the state shown in FIG. 1 as 0 degrees. It is rotatable.

The external view of FIG. 1 is merely an example and is not limited to this external view. If the computer used as the video conferencing terminal 10 is not equipped with a microphone or a camera, an external microphone and a camera can be connected to the computer. When the video conferencing terminal 10 is a general-purpose computer, a smartphone, or the like, the video conferencing terminal 10 may be connected to the Internet by wireless communication via a wireless LAN, a mobile phone network, or the like. When a general-purpose computer is used as the video conferencing terminal 10, an application for executing the processing of the video conferencing terminal 10 can be installed on the computer.

<Hardware configuration of the video conferencing terminal according to the embodiment>
Next, FIG. 2 is a diagram illustrating an example of the hardware configuration of the video conferencing terminal 10. As shown in FIG. 2, the video conferencing terminal 10 includes a CPU (Central Processing Unit) 701, a ROM (Read Only Memory) 702, a RAM (Random Access Memory) 703, a flash memory 704, and an SSD (Solid State Drive). ) 705, a media I / F (Interface) 707, an operation button 708, and a power switch 709. The video conferencing terminal 10 includes a bus line 710, a network I / F 711, a camera 712, an image pickup element I / F 713, a microphone 714, a speaker 715, a sound input / output I / F 716, and a display I / F 717. It has an external device connection I / F 718, a short-range communication circuit 719, and an antenna 719a of the short-range communication circuit 719.

Of these, the CPU 701 controls the operation of the entire video conferencing terminal 10. The ROM 702 stores a program used to drive the CPU 701 such as an IPL (Initial Program Loader). The SSD 704 is used as a work area for the CPU 701. The flash memory 704 stores various data such as a communication program, image data, and sound data.

The SSD 705 controls reading or writing of various data to the flash memory 704 according to the control of the CPU 701. An HDD (Hard Disk Drive) may be used instead of the SSD. The media I / F 707 controls reading or writing (storage) of data to a recording medium 706 such as a flash memory.

The operation button 708 includes a plurality of operation buttons (108a to 108e) shown in FIG. 1 and is operated when selecting a destination of the video conferencing terminal 10. However, the operation of the video conferencing terminal 10 is not limited to the operation buttons 708, and is composed of a liquid crystal display (LCD) and an organic EL (Organic Electro-Luminescence) display device equipped with a touch panel function. May be done.

The power switch 709 is a switch for switching the power ON / OFF of the video conferencing terminal 10. Further, the network I / F711 is an interface for data communication using a communication network such as the Internet.

The CMOS (Complementary Metal Oxide Semiconductor) sensor 712 is a kind of built-in imaging means for capturing an image of a subject and obtaining image data under the control of the CPU 701. Instead of a CMOS sensor, an imaging means such as a CCD (Charge Coupled Device) sensor may be used. The image sensor I / F 713 is a circuit that controls the drive of the CMOS sensor 712. The microphone 714 is a built-in circuit that converts sound into an electrical signal. The speaker 715 is a built-in circuit that converts an electric signal into physical vibration to produce sounds such as music and voice.

The sound input / output I / F 716 is a circuit that processes sound signal input / output between the microphone 714 and the speaker 715 under the control of the CPU 701. The display I / F 717 is a circuit that transmits image data to an external display under the control of the CPU 701.

The external device connection I / F718 is an interface for connecting various external devices. External devices such as an external camera, an external microphone, and an external speaker can be connected to the external device connection I / F718 by a USB (Universal Serial Bus) cable or the like. When an external camera is connected, the external camera is driven in preference to the built-in CMOS sensor 712 according to the control of the CPU 701. Similarly, when an external microphone is connected or an external speaker is connected, the external microphone and the external speaker are given priority over the built-in microphone 714 and the built-in speaker 715 according to the control of the CPU 701. The external speaker is driven.

The short-range communication circuit 719 is a communication circuit such as NFC (Near Field Communication) or Bluetooth (registered trademark). Further, the bus line 710 is an address bus, a data bus, or the like for electrically connecting each component such as the CPU 701 shown in FIG.

The display 720 is a kind of display means composed of a liquid crystal, an organic EL (Electro Luminescence), or the like for displaying an image of a subject, an operation icon, or the like. Further, the display 720 is connected to the display I / F 717 by a cable. This cable may be a cable for analog RGB (VGA) signals, a cable for component video, HDMI (High-Definition Multimedia Interface) (registered trademark), or DVI (Digital Video Interactive). ) It may be a signal cable.

Further, the recording medium 706 has a structure that can be attached to and detached from the video conferencing terminal 10. Further, as long as it is a non-volatile memory that reads or writes data under the control of the CPU 701, not only the flash memory 704 but also an EEPROM or the like may be used.

<Functional configuration of the video terminal according to the embodiment>
Here, when a video conference using a moving image is performed using the video conference terminal 10, the face area including the face of a person in the moving image can convey the facial expression of the conference participant to the other party. This is an area of high importance for facilitating communication. On the other hand, the non-face region that does not include the face is a region that is less important than the region that includes the face.

In the embodiment, in an image such as a moving image, a process of reducing the amount of information is executed only in a non-face region having low importance. As a result, the quality of the image in the face region, which is highly important, is maintained, the amount of information in the image is reduced, and the quality of the image and the sound is prevented from being deteriorated.

FIG. 3 is a block diagram illustrating an example of the functional configuration of the video conferencing terminal 10 according to the embodiment. As shown in FIG. 3, the video conferencing terminal 10 has a face detection processing unit 301, a non-face area setting unit 302, an information amount reduction processing unit 303, and a coding processing unit 304. Each of these parts is realized by the CPU 701 of FIG. 2 executing a predetermined program or the like.

The face detection processing unit 301 detects the face of a person included in the image (input image) of each frame of the moving image input to the video conferencing terminal 10 at a predetermined frame rate. Then, the coordinate information indicating the face area corresponding to the face of the person in the image is output to the non-face area setting unit 302. Further, the face detection processing unit 301 detects the amount of movement of the person's face between the frames of the input moving image and outputs it to the frame rate reduction processing unit 307.

The non-face area setting unit 302 sets an area other than the face area as a non-face area based on the input coordinate information indicating the face area, and outputs an image in which the non-face area is set to the information amount reduction processing unit 303. To do.

The information amount reduction processing unit 303 executes a process of reducing the amount of information for the non-face area set by the image input from the non-face area setting unit 302. Further, the information amount reduction processing unit 303 includes a low-pass filter processing unit 305, a contrast reduction processing unit 306, and a frame rate reduction processing unit 307.

The low-pass filter processing unit 305 selectively executes the low-pass filter processing on the non-face region of the input image. The low-pass filter processing is a so-called low-pass filter processing in which only a low spatial frequency band in an image is passed and other bands are blocked.

Here, the processing performed by the low-pass filter processing unit 305 for the non-face region will be described with reference to FIG. FIG. 4 shows an image 40 including a state in which three people hold a meeting around a desk 41 in a meeting room. In the image 40, the three regions 42 shown by the broken lines are face regions including the face of a person. The area other than the area 42 in the image 40 is a non-face area that does not include a person's face. The low-pass filter processing unit 305 selectively executes the low-pass filtering process on such a non-face region. Similarly, the contrast reduction processing unit 306 and the frame rate reduction processing unit 307 described below also selectively execute processing on such a non-face region.

Returning to FIG. 3, the description of FIG. 3 is continued. The contrast reduction processing unit 306 executes a process of selectively reducing the contrast in the non-face region. The contrast is the difference between light and dark in an image, and the contrast reduction processing unit 306 performs a process of reducing this, that is, reducing the difference between light and dark.

The frame rate reduction processing unit 307 executes a process of selectively reducing the frame rate for the non-face region of the input image. The frame rate is the number of frames processed per unit time in a moving image, that is, the number of still images or the number of frames. As an example, when the frame rate of a moving image is set to 30 fps (frame per Second), the frame rate reduction processing unit 307 selectively reduces the frame rate to 15 fps or the like with respect to the non-face region.

Here, the processing by the frame rate reduction processing unit 307 will be described with reference to FIG. FIG. 5 is a timing chart showing an example of update timing for each image region in a moving image.

In FIG. 5, the signal FR is a signal indicating the timing at which one frame of the moving image is input to the video conferencing terminal 10. In FIG. 5, four-frame images are input to the video conferencing terminal 10 at the timing indicated by the signal FR, respectively.

The signal AR1 is a signal indicating the timing at which the face region of the image of one frame in the moving image is updated. The signal AR2 is also a signal indicating the timing at which the non-face region is updated. In the signal AR1, the image of the face area is updated once every time the image of one frame is input once. On the other hand, in the signal AR2, the image in the non-face region is updated once every time the image of one frame is input twice. That is, in the non-face region, the frame rate is lowered with respect to the face region by thinning out the update once. In this way, the frame rate reduction processing unit 307 can selectively reduce the frame rate of the non-face region of the input image.

Returning to FIG. 3, the description of FIG. 3 is continued. The information amount reduction processing unit 303 outputs an image in which the amount of information is selectively reduced with respect to the non-face region of the input image to the coding processing unit 304. In the embodiment, examples are shown in which the low-pass filter processing unit 305, the contrast reduction processing unit 306, and the frame rate reduction processing unit 307 are all executed, but the present invention is not limited thereto. The information amount reduction processing unit 303 has at least one of a low-pass filter processing unit 305, a contrast reduction processing unit 306, or a frame rate reduction processing unit 307, so that at least one processing by each unit is executed. Is also good.

The coding processing unit 304 encodes the image input from the information amount reduction processing unit 303. The coding processing unit 304 is, for example, H.I., which is one of the moving image compression standards. The coding process based on 264 is executed. The encoded image is sequentially output from the video conferencing terminal 10 to an external device as a one-frame image in the moving image. The external device is a video conferencing terminal or the like of the other party performing the video conferencing, and the image encoded by the coding processing unit 304 is output to the video conferencing terminal of the other party via the network.

Next, the effect of reducing the amount of information in the image by each of the low-pass filter processing unit 305 and the contrast reduction processing unit 306 will be described with reference to FIG.

FIG. 6 is a diagram illustrating the effect of reducing the amount of information in the image by the video conferencing terminal 10. In FIG. 6, the vertical axis shows the amount of information in the image. However, the vertical axis is displayed on a standardized scale with the amount of information of the image when no processing is performed on the non-face region as 1. The horizontal axis shows the classification representing each process. In FIG. 6, classifications 61-65 are shown, and eight bar graphs are shown for each classification. The eight bar graphs show the processing results for eight types of test images.

Classification 61 is a case where no processing is performed on the non-face region. Classification 62 is a case where the non-face region is processed by the low-pass filter processing unit 305. Classification 63 is a case where a process of reducing the number of colors is executed in the non-face region as a first comparative example. Classification 64 is a case where the non-face region is processed by the contrast reduction processing unit 306. Classification 65 is a case where noise removal processing is executed on a non-face region as a second comparative example.

“1/2”, “1/4”, and “1/8” in the classification 62 indicate the level at which the resolution of the image is lowered by the amount that the high spatial frequency band is blocked by the low-pass filter processing. .. “1/2”, “1/4”, and “1/8” in the classification 63 indicate the level at which the number of colors in the image is reduced by the reduction processing of the number of colors. Further, "1/2", "1/4", and "1/8" in the classification 64 indicate the level at which the contrast of the image is reduced by the contrast reduction processing.

As described above, since the processing of classification 61 is used as the standard for standardization, in classification 61, the amount of image information is 1 for all eight bar graphs. In classification 62, the amount of information in the image is greatly reduced as the resolution is lowered. In classification 63, the effect of reducing the amount of information in the image due to the decrease in the number of colors is not observed, but rather the amount of information increases in accordance with the decrease in the number of colors. It is considered that this is because the tone jump increased due to the decrease in the number of colors and the compression effect was hindered. In classification 64, the amount of information in the image decreases as the contrast of the resolution decreases. In classification 65, the amount of information in the image is slightly reduced due to noise removal.

As can be seen from FIG. 6, the low-pass filter processing by the low-pass filter processing unit 305 and the contrast reduction processing by the contrast reduction processing unit 306 are more preferably performed in comparison with the first and second comparative examples. The amount of information can be reduced.

Next, an example of processing by the video conferencing terminal 10 according to the embodiment will be described with reference to the flowchart of FIG. 7.

First, in step S71, the face detection processing unit 301 inputs an image of one frame of the moving images.

Subsequently, in step S72, the face detection processing unit 301 executes face detection processing on the input image and outputs the coordinate information of the detected face area to the non-face area setting unit 302.

Subsequently, in step S73, the non-face area setting unit 302 sets the non-face area in the input image, and outputs the set image to the information amount reduction processing unit 303.

Subsequently, in step S74, the low-pass filter processing unit 305 included in the information amount reduction processing unit 303 selectively executes the low-pass filter processing on the non-face region to reduce the contrast of the processed image. Output to the processing unit 306.

Subsequently, in step S75, the contrast reduction processing unit 306 selectively executes the contrast reduction processing on the non-face region, and outputs the processed image to the frame rate reduction processing unit 307.

Subsequently, in step S76, the frame rate reduction processing unit 307 determines whether or not the amount of movement of the face detected by the face detection processing unit 301 is equal to or less than a predetermined movement threshold value. As an example, the face detection processing unit 301 includes the sum of the pixel brightness of the face area in the image of the previous one frame stored in the RAM 703 or the like and the pixel brightness of the face area in the image of the one frame currently being processed. By comparing the sum of the above, the amount of facial movement can be detected.

When it is determined in step S76 that the amount of movement of the face is equal to or less than the movement threshold value (step S76, Yes), in step S77, the frame rate reduction processing unit 307 selectively reduces the frame rate with respect to the non-face region. Execute the process. On the other hand, if it is determined in step S76 that the amount of movement of the face is not equal to or less than the movement threshold value (steps S76, No), the process proceeds to step S78. The movement threshold value is obtained in advance and stored in SSD 704 or the like.

On the other hand, the face detection processing unit 301 stores the current one-frame image in the RAM 703 or the like in order to use it for detecting the amount of movement of the face in the next one-frame image.

Here, the detection of the amount of facial movement is supplemented. When the change between frames in the moving image is small, even if the frame rate reduction processing is selectively executed for the non-face region, the moving image after the processing does not become unnatural. However, if the frame rate reduction processing is selectively executed for the non-face region when the change between frames is large, the change between frames is awkwardly visually recognized, and the processed moving image becomes unnatural. May become.

In the embodiment, the frame rate reduction processing is selectively executed for the non-face region of the input image only when the detected facial movement amount is equal to or less than the movement threshold value, and the facial movement amount (change between frames). By not lowering the frame rate for an image with a large image, it is possible to prevent the moving image from becoming unnatural.

The order of the low-pass filter processing in step S74, the contrast reduction processing in step S75, and the frame rate reduction processing in steps S76 to S77 may be changed as appropriate.

Subsequently, in step S78, the coding processing unit 304 executes the coding processing on the image input from the information amount reduction processing unit 303.

Subsequently, in step S79, the coding processing unit 304 outputs the coded image to the external device as a one-frame image in the moving image.

In this way, it is possible to reduce the amount of information such as an unimportant background while maintaining the amount of information on the face of a person who is highly important in the image.

As described above, according to the embodiment, it is possible to reduce the amount of information in the image and increase the compression rate of the image while maintaining the subjective image quality. Thereby, for example, when the moving image is transmitted using the Internet line, it is possible to prevent the quality of the moving image and the sound from being deteriorated. Further, even when a large amount of moving images are stored like a surveillance camera, the storage capacity can be reduced, which is preferable.

On the other hand, as a method of increasing the compression rate of the image, it is also conceivable to reduce the noise signal of the image. However, the reduction of the noise signal is effective in the case of video conferencing using a projector or the like in a dark room, but it is large in the case of video conferencing in a bright room because the noise signal itself of the image becomes small in the first place. No effect is obtained. In recent years, a backlight type display device such as a liquid crystal display has been used, and video conferencing is often performed in a bright room. Therefore, it is difficult to improve the image compression rate by reducing noise signals. The video conferencing terminal 10 according to the embodiment can further increase the compression rate of the image as compared with the method of reducing such a noise signal.

[First Embodiment]
Next, the video conferencing terminal 10a according to the first embodiment will be described. The description of the same components as those of the above-described embodiment will be omitted.

As described above, the non-face region that does not include the human face is less important than the face region, and therefore it is preferable to reduce the amount of information as much as possible. However, when the image characteristics (feature amount) such as brightness are significantly different between the face area and the non-face area, if the amount of information between the face area and the non-face area is significantly different, the boundary between the areas becomes conspicuous and the feeling of strangeness is felt. It may result in some low quality image.

Therefore, in the present embodiment, when the feature amount such as brightness is significantly different between the face area and the non-face area, the amount of information in the non-face area is not sharply reduced with respect to the face area, and as the distance from the face area increases. By gradually reducing the amount of information, deterioration of the quality of the processed image is suppressed.

FIG. 8 is a block diagram illustrating an example of the functional configuration of the video conferencing terminal 10a according to the first embodiment. As shown in FIG. 8, the video conferencing terminal 10a has a first partition processing unit 308 and an information amount reduction processing unit 303a.

The first division processing unit 308 executes a process of dividing the non-face region of the image input from the non-face region setting unit 302 into a plurality of frame-shaped regions surrounding the face. For this division process, first division information such as the number of divisions N and the distance D from the face area is predetermined and stored in SSD 704 or the like in FIG. Further, the first division processing unit 308 includes a frame-shaped region setting unit 309, a first feature amount extraction unit 310, and a first division determination unit 311.

The frame-shaped area setting unit 309 sets a plurality of rectangular frame-shaped areas so as to surround the detected face according to the first division information stored in the SSD 704 or the like. Here, FIG. 9 is a diagram illustrating an example of a frame-shaped region set by the frame-shaped region setting unit 309. The directions indicated by the white arrows in FIG. 9 indicate the X direction corresponding to the X coordinate of the image and the Y direction corresponding to the Y coordinate, respectively.

In FIG. 9, the face areas 42a and 42b are areas corresponding to the faces of two persons detected by the face detection processing unit 301. Rectangular frame-shaped regions 91a, 92a and 93a are set around the face region 42a, respectively, and rectangular frame-shaped regions 91b, 92b and 93b are set around the face region 42b, respectively.

The frame-shaped region 91a is a region separated from the center position of the face region 42a in the X and Y directions by a distance D1 or more and a distance less than D2, respectively. The unit of distance is the number of pixels. The frame-shaped region 92a is a region separated from the center position of the face region 42a in the X and Y directions by a distance D2 or more and a distance less than D3, respectively. The frame-shaped region 93a is a region separated from the center position of the face region 42a in the X and Y directions by a distance D3 or more, respectively.

Further, the frame-shaped region 91b is a region separated from the center position of the face region 42b in the X and Y directions at a distance of D1 or more and less than D2, respectively. The frame-shaped region 92b is a region separated from the center position of the face region 42b in the X and Y directions at a distance of D2 or more and less than D3, respectively. The frame-shaped region 93b is a region separated from the center position of the face region 42b in the X and Y directions by a distance D3 or more, respectively.

Here, in the corresponding image region between the face region 42a and the face region 42b, the frame-shaped region may overlap. For example, in FIG. 9, the negative X-direction side (face region 42b side) of the face region 42a in the frame-shaped region 93 is the positive X-direction side (face region 42a side) of the face region 42b in the frame-shaped region 91b. It overlaps with the area of. The region where the frame-shaped regions overlap in this way is set as a region belonging to both frame-shaped regions. On the other hand, when the frame-shaped area overlaps with the face area, the overlapping area is excluded from the frame-shaped area.

However, the setting for the overlapping area is not limited to the above-mentioned one, and it is possible to set the setting rule for the overlapping area in advance and set according to the rule.

After setting the rectangular frame-shaped area as shown in FIG. 9, the first feature amount extraction unit 310 calculates the average value of the brightness of all the pixels included in the face area and each frame-shaped area in each of the face area and each frame-shaped area. , Calculate the difference in the average brightness between adjacent frame-shaped regions and between adjacent face regions and frame-shaped regions. Here, the difference in the average value of the brightness of all the pixels included in the region is an example of the "first feature amount".

More specifically, the first feature amount extraction unit 310 has an average luminance value Ave42a of all the pixels included in the face region 42a, an average luminance value Ave91a of all the pixels included in the frame-shaped region 91a, and a frame-shaped region 92a. The average value Ave92a of the brightness of all the pixels included in the above and the average value Ave93a of the brightness of all the pixels included in the frame-shaped region 93a are calculated respectively. In addition, the first feature amount extraction unit 310 includes an average value Ave42b of the brightness of all the pixels included in the face area 42b, an average value Ave91b of the brightness of all the pixels included in the frame-shaped area 91b, and all included in the frame-shaped area 92b. The average value Ave92b of the brightness of the pixels and the average value Ave93b of the brightness of all the pixels included in the frame-shaped region 93b are calculated, respectively.

Then, the first feature amount extraction unit 310 has "Ave42a-Ave91a", "Ave91a-Ave92a", "Ave92a-Ave93a", "Ave42b-Ave91b", "Ave91b-Ave92b", and "Ave92b-" as the difference between the average values. "Ave93b" is calculated respectively, and the calculation result is output to the first division determination unit 311.

After that, the first division determination unit 311 compares each of the calculated differences in the average values of the brightness with the predetermined first brightness threshold value, and determines whether or not the difference in the average value is larger than the first brightness threshold value. judge. When it is determined that the difference between the average values of the luminances is larger than the first luminance threshold value, the first division determination unit 311 divides the adjacent regions into two regions. On the other hand, when the average value of the brightness is equal to or less than the first brightness threshold value, the first division determination unit 311 does not divide the adjacent regions, but combines the two to form one frame-shaped region. Here, the first luminance threshold is an example of the "first threshold".

As an example, the difference between the average values "Ave42a-Ave91a", "Ave91a-Ave92a", "Ave92a-Ave93a", and "Ave42b-Ave91b" are larger than the first luminance threshold value, and the difference between the average values "Ave91b-Ave92b" and "Ave91b-Ave92b". When "Ave92b-Ave93b" is smaller than the first luminance threshold value, the non-face region is divided as shown in FIG. FIG. 10 is a diagram illustrating an example of a non-face region divided by the first division processing unit 308.

In FIG. 10, since the difference between the average values “Ave42a-Ave91a”, “Ave91a-Ave92a”, “Ave92a-Ave93a” and “Ave42b-Ave91b” is larger than the first luminance threshold value, the adjacent face region 42a and the frame-shaped region 91a , The adjacent frame-shaped region 91a and the frame-shaped region 92a, the adjacent frame-shaped region 92a and the frame-shaped region 93a, and the adjacent face region 42b and the frame-shaped region 91b are each divided. The divided frame-shaped regions are treated as separate regions, and the information amount reduction processing unit 303a executes different processing for each region.

On the other hand, since the differences "Ave91b-Ave92b" and "Ave92b-Ave93b" are equal to or less than the first luminance threshold value, the frame-shaped regions 91b, 92b and 93b are not divided and are combined into one frame-shaped region 93b. .. A common process is executed in the frame-shaped region 93b by the information amount reduction processing unit 303a.

Here, as described above, the frame-shaped region may overlap in the corresponding image region between the face region 42a and the face region 42b. In the area where the frame-shaped areas overlap, when it is determined that one is not divided and the other is determined to be divided, the process of dividing is prioritized. However, the setting for the overlapping area is not limited to the above-mentioned one, and the setting rule for the overlapping area can be set in advance and the division process can be performed according to the rule.

The low-pass filter processing unit 305a included in the information amount reduction processing unit 303a of FIG. 8 uses a spatial filter having a different tap size for each frame-shaped region of the non-face region divided by the first division processing unit 308. , Perform spatial filtering.

11A and 11B are views for explaining an example of a spatial filter used in the video conferencing terminal 10a, FIG. 11A is a diagram for explaining a spatial filter 101 having a size of 3 × 3 pixels, and FIG. 11B is a diagram for explaining a spatial filter 101 having a size of 5 × 5 pixels. It is a figure explaining the space filter 102 of.

The low-pass filter processing unit 305a executes a convolution (convolution integration) process between the spatial filter 101 and each pixel in the frame-shaped region. By integrating and adding the values shown in each square of the spatial filter 101 to each pixel in the frame-shaped region, the difference in brightness between the adjacent pixels in the frame-shaped region is reduced. As a result, the high frequency band of the spatial frequency in the image of the frame-shaped region is cut off, and the effect of the low-pass filter (low-pass filter) can be obtained. The same applies to the spatial filter 102.

The larger the tap size, the smaller the difference in brightness between adjacent pixels in the frame-shaped region, the wider the frequency band blocked on the high frequency side, and the greater the effect of reducing the amount of information by the low-pass filter. Therefore, the effect of reducing the amount of information is greater in the space filter 102 with 5 × 5 pixels than in the space filter 101 with 3 × 3 pixels.

On the other hand, the larger the difference in tap size between the adjacent frame-shaped regions or between the adjacent face region and the frame-shaped region, the more conspicuous the boundary between the regions becomes after the low-pass filter processing.

The low-pass filter processing unit 305a executes low-pass filter processing on the frame-shaped region 91a in FIG. 10 by using the space filter 101, and uses the space filter 102 on the frame-shaped region 92a. And low-pass filter processing is executed. Further, for the frame-shaped region 93a, the low-pass filter processing is executed by using a 7 × 7 pixel spatial filter or the like having a tap size larger than that of the spatial filter 102.

Here, in the face region 42a, the low-pass filter processing is not executed in order to maintain the amount of information. Therefore, when the difference in the average value of the pixel brightness between the face region 42a and the non-face region is larger than the first luminance threshold value, a 7 × 7 pixel spatial filter having a large tap size is applied to the entire non-face region. When the low-pass filtering process is performed using the method, the boundary between the face region 42a and the non-face region may be conspicuous.

On the other hand, the low-pass filter processing unit 305a executes the low-pass filter processing in the frame-shaped region 91a adjacent to the face region 42a by using a spatial filter of 3 × 3 pixels having a small tap size. Further, in the frame-shaped region 92a adjacent to the frame-shaped region 91a, a low-pass filter process is executed by using a space filter having 5 × 5 pixels, which is close to a space filter having 3 × 3 pixels. Similarly, in the frame-shaped region 93a adjacent to the frame-shaped region 92a, a low-pass filter process is executed using a 7 × 7 pixel spatial filter close to a 5 × 5 pixel spatial filter. By doing so, it is possible to prevent the tap size of the spatial filter from being significantly different between adjacent areas. As a result, the boundaries between the regions can be made inconspicuous after the low-pass filter processing.

On the other hand, around the face region 42b, the frame-shaped regions 91b, 92b and 93b are combined and treated as one frame-shaped region 93b. When combining the frame-shaped areas, a tap-sized spatial filter applied to the frame-shaped area on the side far from the face area is adopted. Therefore, in the frame-shaped region 93b after coalescence, the 7 × 7 pixel tap size spatial filter used in the farthest frame-shaped region among the tap sizes applied to the frame-shaped regions 91b, 92b and 93b is used. Used for low-pass filtering.

The difference in the average value of the pixel luminance between the face region 42b and the frame-shaped region 93b after coalescence is smaller than the first luminance threshold and is small. Therefore, even if the low-pass filter processing is executed using the 7 × 7 pixel spatial filter having a large tap size on the entire frame-shaped region 93b, the boundary between the face region 42b and the non-face region is conspicuous. There is no such thing. Therefore, a large amount of information can be reduced by using a spatial filter having a large tap size.

In this way, when the process of reducing the amount of information is executed, the amount of information in the image can be reduced while maintaining the quality of the image by making the boundaries between the regions inconspicuous.

<Processing by the video conferencing terminal according to the first embodiment>
FIG. 12 is a flowchart showing an example of processing by the video conferencing terminal 10a according to the present embodiment.

Here, since the processing of steps S121 to S123 is the same as the processing of steps S71 to S73 of FIG. 7, duplicate description will be omitted here.

In step S124, the frame-shaped area setting unit 309 sets a plurality of rectangular frame-shaped areas so as to surround the detected face according to the first division information stored in the SSD 704 or the like.

Subsequently, in step S125, the first feature amount extraction unit 310 calculates the average value of the brightness of all the pixels included in the face region and each frame-shaped region in each of the face region and each frame-shaped region, and the adjacent frame-shaped regions are connected to each other. Calculate the difference in the average brightness between the face area and the adjacent face area and the frame-shaped area. After that, the calculation result is output to the first division determination unit 311.

Subsequently, in step S126, the first division determination unit 311 compares each of the differences in the average luminance values input from the first feature quantity extraction unit 310 with the first luminance threshold value, and the difference in the average luminance values is the difference. It is determined whether or not it is larger than the first luminance threshold.

When it is determined in step S126 that the difference between the average values of the brightness is larger than the first brightness threshold value (step S126, Yes), in step S127, the first division determination unit 311 divides the adjacent region into two regions. To do.

On the other hand, when it is determined in step S126 that the difference between the average luminance values is equal to or less than the first luminance threshold value (step S126, No), in step S128, the first division determination unit 311 does not divide the adjacent region. , Both are combined into one frame-shaped area.

Such processing of steps S126 to S128 is executed for all of the adjacent frame-shaped regions and the adjacent face region and frame-shaped region.

Subsequently, in step S129, the low-pass filter processing unit 305a executes the spatial filter processing by using a spatial filter having a different tap size for each frame-shaped region divided by the first division processing unit 308.

Since the processes of steps S130 to S134 are the same as the processes of steps S75 to S79 of FIG. 7, redundant description will be omitted here.

In steps S130 to S132, an example of executing a common contrast reduction process and a frame rate reduction process for the entire non-face region has been shown, but the frame of the non-face region divided by the first division processing unit 308 has been shown. Different contrast reduction processing and frame rate reduction processing may be executed for each state region.

In this way, the video conferencing terminal 10a can divide the non-face region into a plurality of frame-shaped regions and execute a process of reducing the amount of information for each frame-shaped region.

<Effect>
As described above, in the present embodiment, the non-face region of the input image to the video conferencing terminal 10a is divided into a plurality of frame-shaped regions based on the brightness of the pixels included in the region. Specifically, in the non-face region, the difference in the average brightness of all the pixels included in each region between the adjacent frame-shaped regions and between the adjacent face region and the frame-shaped region is calculated. calculate. Then, when the difference between the average values of the brightness is larger than the first brightness threshold value, the adjacent frame-shaped regions or the adjacent face region and the frame-shaped region are divided. Further, when the difference between the average values of the brightness is equal to or less than the first brightness threshold value, the adjacent frame-shaped regions or the adjacent face region and the frame-shaped region are combined without being divided into one region. After that, the amount of information is reduced by using a spatial filter having a different tap size for each divided region of the non-face region.

By doing so, when the difference in the average brightness between the adjacent frame-shaped regions or between the adjacent face region and the frame-shaped region is large, the spatial filter used in the low-pass filter processing can be used. It is possible to prevent the tap size from being significantly different between adjacent areas. As a result, the boundaries between the regions can be made inconspicuous after the low-pass filter processing, and the quality of the image can be maintained.

In addition, it is possible to reduce the amount of information such as an unimportant background while maintaining the amount of information of a person who is important information in an image. In this way, the amount of information in the image can be reduced and the compression rate of the image can be increased while maintaining the subjective image quality.

In the above-mentioned example, a rectangular frame-shaped region has been described as an example of the frame-shaped region, but a frame-shaped region having another shape such as a circle may be used. Further, although the convolution processing using the spatial filter has been described as an example of the low-pass filter processing, other processing for lowering the spatial frequency such as the Fourier transform processing can also be applied.

Further, as an example of the first feature amount, the difference in the brightness average value of all the pixels included in the region between adjacent regions has been described, but the maximum value, the minimum value, and the total of the brightness of all the pixels included in the region are described. , The difference in contrast (difference between the maximum value and the minimum value of the brightness) between adjacent regions may be used.

It should be noted that the difference determination of the feature amount of the divided frame-shaped area may be held in a memory such as SSD 704, and it may be determined whether to divide the frame-shaped area according to the threshold value of the N number of the difference determination set separately. .. By doing so, it is possible to make a finer judgment as compared with the case of determining whether to divide the frame-shaped region based on the feature amount obtained from the entire frame.

[Second Embodiment]
Next, the video conferencing terminal 10b according to the second embodiment will be described.

In the frame-shaped region divided by the first division processing unit 308 described in the first embodiment, image characteristics (features) such as brightness may be partially different. In this case, if the same processing is executed for the entire frame-shaped area in order to reduce the amount of information, the difference in the feature amount such as brightness is conspicuous for each part in the frame-shaped area, which makes the person feel uncomfortable. It may result in some low quality image. Therefore, in the present embodiment, when the feature amount such as brightness is significantly different for each part in the frame-shaped area, the frame-shaped area is divided into a plurality of rectangular areas, and the amount of information is reduced for each rectangular area. By making them different, deterioration of the quality of the processed image is suppressed.

<Functional configuration of the video conferencing terminal according to the second embodiment>
FIG. 13 is a diagram illustrating an example of the functional configuration of the video conferencing terminal 10b according to the present embodiment. As shown in FIG. 13, the video conferencing terminal 10b has a second division processing unit 312 and an information amount reduction processing unit 303b. These are realized by the CPU 701 of FIG. 2 executing a predetermined program or the like.

The second division processing unit 312 executes a process of dividing the frame-shaped region divided by the first division processing unit 308 into a plurality of rectangular regions based on the second feature amount of the input image. For this division process, second division information such as the number of divisions P and the size Q of the rectangular area is predetermined and stored in SSD 704 or the like. In addition, the second division processing unit 312 has a rectangular area setting unit 313, a second feature amount extraction unit 314, and a second division determination unit 315.

The rectangular area setting unit 313 sets a plurality of rectangular areas at the boundary of the frame-shaped area for each of the frame-shaped areas divided by the first division processing unit 308 according to the second division information stored in the SSD 704 or the like. To do. Here, FIG. 14 is a diagram illustrating an example of a rectangular area set by the rectangular area setting unit 313.

In FIG. 14, the twelve rectangular regions 94a shown by diagonal hatching are rectangular regions set at the boundaries of the frame-shaped regions with respect to the frame-shaped regions 92a divided by the first division processing unit 308. Note that FIG. 14 shows an example in which the rectangular area is set only for the frame-shaped area 92a in order to simplify the drawing, but the rectangular area setting unit 313 is divided by the first partition processing unit 308. For all frame-shaped areas, a rectangular area can be set at the boundary of the frame-shaped area.

After setting the rectangular region 94a in the frame-shaped region as shown in FIG. 14, the second feature amount extraction unit 314 sets the average value of the brightness of all the pixels included in the region in each of the face region and each rectangular region. calculate. Then, the difference in the average value of the brightness of all the pixels included in the area is calculated between the adjacent rectangular areas and between the adjacent face area and the rectangular area, and the calculation result is determined by the second division determination unit. Output to 315. Here, the average value of the brightness of all the pixels included in the area is an example of the "second feature amount".

After that, the second division determination unit 315 compares each of the calculated differences in the average values of the brightness with a predetermined second brightness threshold. Then, when the difference between the average values of the brightness is larger than the second brightness threshold value, the adjacent regions are divided into two regions. On the other hand, when the average value of the brightness is equal to or less than the second brightness threshold value, the adjacent regions are not divided but combined into one region. Here, the second luminance threshold is an example of the "second threshold". The second threshold value may be the same as or different from the first threshold value described above.

Each of the divided regions is treated as a separate region, and the information amount reduction processing unit 303b executes different processing for each region. On the other hand, the combined regions that are not divided are treated as one region, and the information amount reduction processing unit 303b executes common processing.

The low-pass filter processing unit 305b included in the information amount reduction processing unit 303b uses a spatial filter having a different tap size for each rectangular area 94a in the frame-shaped area of the non-face area divided by the second division processing unit 312. And execute spatial filtering. Here, in the present embodiment, when the tap size of the spatial filter is different, the tap size is not significantly different between the adjacent rectangular areas in order to make the boundary inconspicuous. For example, when applying a 3 × 3 pixel tap size spatial filter to one adjacent rectangular area, tap the other rectangular area for 3 × 3 pixels such as a 5 × 5 pixel tap size spatial filter. Use spatial filters that are close in size.

<Processing by the video conferencing terminal according to the second embodiment>
FIG. 15 is a flowchart showing an example of processing by the video conferencing terminal 10b according to the present embodiment.

Here, since the processing of steps S151 to S154 is the same as the processing of steps S121 to S124 of FIG. 12, duplicate description will be omitted here.

In step S155, the rectangular area setting unit 313 has a plurality of rectangular areas at the boundary of the frame-shaped area with respect to the frame-shaped area divided by the first division processing unit 308 according to the second division information stored in the SSD 704 or the like. To set.

Subsequently, in step S156, the second feature amount extraction unit 314 calculates the average value of the brightness of all the pixels included in each of the face region and each rectangular region. Then, the difference in the average value of the brightness between the adjacent rectangular regions and between the adjacent face region and the rectangular region is calculated, and the calculation result is output to the second division determination unit 315.

Subsequently, in step S157, the second division determination unit 315 compares each of the differences in the average luminance values input from the second feature quantity extraction unit 314 with the second luminance threshold value, and the difference in the average luminance values is the second. 2 Determine whether or not it is larger than the brightness threshold.

When it is determined in step S157 that the difference between the average values of the luminances is larger than the second luminance threshold value (steps S157, Yes), in step S158, the second division determination unit 315 divides the adjacent region into two regions. To do.

On the other hand, when it is determined in step S157 that the difference between the average values of the brightness is equal to or less than the second brightness threshold value (step S157, No), in step S159, the second division determination unit 315 does not divide the adjacent region. The two are combined into one frame-shaped area.

Subsequently, in step S160, the low-pass filter processing unit 305b executes spatial filter processing using a spatial filter having a different tap size for each rectangular region of the non-face region divided by the second division processing unit 312. To do.

Since the processes of steps S161 to S165 are the same as the processes of steps S130 to S134 of FIG. 12, redundant description will be omitted here.

In steps S161 to S165, an example of executing a common contrast reduction process and a frame rate reduction process for the entire non-face region has been shown, but the rectangle of the non-face region divided by the second division processing unit 312 has been shown. The contrast reduction process and the frame rate reduction process, which are different for each region, may be executed respectively.

In this way, the video conferencing terminal 10b can divide the frame-shaped region of the non-face region into a plurality of rectangular regions and execute a process of reducing the amount of information for each rectangular region.

<Effect>
As described above, in the present embodiment, the non-face region of the input image to the video conferencing terminal 10b is divided into a plurality of frame-shaped regions based on the brightness of the pixels included in the region, and a plurality of frame-shaped regions are further divided. The area is divided into a plurality of rectangular areas based on the brightness of the pixels included in the area.

Specifically, in the frame-shaped region, the difference in the average value of the brightness of all the pixels included in each region is calculated between the adjacent rectangular regions and between the adjacent face region and the rectangular region. Then, when the difference between the average values of the brightness is larger than the first brightness threshold value, the adjacent rectangular regions or the adjacent face region and the rectangular region are divided. Further, when the difference between the average values of the brightness is equal to or less than the second brightness threshold value, the adjacent rectangular regions or the adjacent face region and the rectangular region are divided. After that, the amount of information is reduced by using spatial filters having different tap sizes for each divided area.

By doing so, when the difference in the average value of the brightness of all the pixels included in each area is large between the adjacent rectangular areas or between the adjacent face area and the rectangular area, the low range is used. It is possible to prevent the tap size of the spatial filter used in the pass filtering process from being significantly different between adjacent regions. As a result, the boundaries between the regions can be made inconspicuous after the low-pass filter processing, and the quality of the image can be maintained.

In addition, it is possible to reduce the amount of information such as an unimportant background while maintaining the amount of information of a person who is important information in an image. In this way, the amount of information in the image can be reduced and the compression rate of the image can be increased while maintaining the subjective image quality.

In the above-mentioned example, as an example of the second feature amount, the difference in the brightness average value of all the pixels included in the region between the adjacent regions has been described, but the maximum value of the brightness of all the pixels included in the region has been described. , The minimum value, the sum, and the difference in contrast (difference between the maximum value and the minimum value of the brightness) between adjacent regions may be used.

Although the image processing apparatus according to the embodiment has been described above, the present invention is not limited to the above embodiment, and various modifications and improvements can be made within the scope of the present invention.

The embodiment also includes a program. For example, the program divides the computer by the first division processing unit that divides the area other than the face area of the person included in the input image based on the first feature amount of the input image, and the first division processing unit. It functions as an information amount reduction processing unit that reduces the amount of information in each of the created areas. With such a program, the same effect as that of the image processing apparatus described above can be obtained.

Further, the embodiment also includes an image processing method. For example, the image processing method includes a step of dividing a region other than the face region of a person included in the input image based on the first feature amount of the input image, and each region divided by the first division processing unit. , And a step of executing a process of reducing the amount of information. By such an image processing method, the same effect as that of the image processing apparatus described above can be obtained.

Further, each function of the embodiment described above can be realized by one or a plurality of processing circuits. Here, the "processing circuit" in the present specification is a processor programmed to execute each function by software such as a processor implemented by an electronic circuit, or a processor designed to execute each function described above. It shall include devices such as ASIC (Application Specific Integrated Circuit), DSP (digital signal processor), FPGA (field programmable gate array) and conventional circuit modules.

10, 10a Video conferencing terminal (an example of image processing device)
40 Images 42a, 42b Face areas 91a to 93a, 91b to 93b Frame-shaped areas 94a Rectangular areas 101, 102 Spatial filter 301 Face detection processing unit 302 Non-face area setting unit 303 Information amount reduction processing unit 304 Coding processing unit 305 Low range Pass filter processing unit 306 Contrast reduction processing unit 307 Frame rate reduction processing unit 308 First division processing unit 309 Frame-shaped area setting unit 310 First feature amount extraction unit 311 First division determination unit 312 Second division processing unit 313 Rectangular area setting Part 314 Second feature amount extraction part 315 Second division determination part

JP-A-2002-165222

Claims (6)

A first division processing unit that divides an area other than the face area of a person included in the input image based on the first feature amount of the input image.
An image processing apparatus including an information amount reduction processing unit that executes a process of reducing the amount of information for each area divided by the first division processing unit. The first feature amount is
Pixel brightness between a plurality of frame-shaped regions surrounding the face region and the face region in a region other than the face region, between adjacent frame-shaped regions and between the adjacent face region and the frame-shaped region. Extracted based on the difference between
The first partition processing unit
The image processing apparatus according to claim 1, wherein when the difference in pixel brightness is larger than a predetermined first threshold value, the adjacent frame-shaped regions or the adjacent face region and the frame-shaped region are divided. It has a second division processing unit that divides the frame-shaped region based on the second feature amount of the input image.
The information amount reduction processing unit
The image processing apparatus according to claim 2, wherein the processing for differentiating the amount of information is executed for each region divided by the second division processing unit. The second feature amount is
Extracted based on the difference in pixel brightness between the adjacent rectangular regions and between the adjacent face regions and the rectangular regions among the plurality of rectangular regions and the face regions in each of the plurality of frame-shaped regions. Being done
The second division processing unit
The image processing apparatus according to claim 3, wherein when the difference in pixel brightness is larger than a predetermined second threshold value, the adjacent rectangular regions or the adjacent face region and the rectangular region are divided. Computer,
Based on the first feature amount of the input image, the amount of information is divided by the first division processing unit that divides the area other than the face area of the person included in the input image and the area divided by the first division processing unit. A program that functions as an information amount reduction processing unit that executes processing that reduces the amount of information. A step of dividing a region other than the human face region included in the input image based on the first feature amount of the input image, and
An image processing method including a step of executing a process of reducing an amount of information for each area divided by the first division processing unit.

Images