JPH0595541A - Picture encoder - Google Patents

Abstract

PURPOSE:To improve the picture quality of the composite picture in the specific area such as a face area and to display the still background in a constant pattern by performing the filter processing in the time axis direction and the space direction in the dynamic area other than the face area. CONSTITUTION:A face area detection means 41 detects the face area in an input picture signal, and outputs the specific area designation signal. A dynamic area detection means 5 detects the dynamic area including the specific area of the input picture and outputs the dynamic area designation signal. A time axis filter 2 and a space filter 3 filter the picture signal of the area other than the specific area in the time axis direction and the picture signal of the dynamic range other than the face area in the time axis direction and the space direction among the input picture signals according to the specific area designation signal and the dynamic area designation signal. An encoder 6 encodes the output picture from the low-pass filter means 2 and 3 by an encoding circuit. Thus, the generation information amount of the area other than the face area is reduced in quantization, and the background will not be blurred.

Description

Detailed Description of the Invention

[0001]

BACKGROUND OF THE INVENTION 1. Field of the Invention The present invention relates to an image coding apparatus for coding a moving image in a video conference, a video telephone or the like.

[0002]

2. Description of the Related Art In recent years, moving image videophones and videoconference systems have been put into practical use due to the development of digital communication networks and the development of communication technologies. In particular, since the video conference system is extremely effective in saving the time and cost required for a business trip, its introduction is being promoted in companies. Video coding is one of the most important technologies for realizing videophones and videoconferencing systems. As one of such encoding methods, a method of detecting a face region from a moving image signal of a human image and filtering image signals other than the face region in the time axis direction and the spatial direction to reduce the amount of information other than the face region. There is.

In this case, the previous frame image and the current frame image are compared with each other, and a portion having a large change is detected as a face area.
This face area is directly encoded without any filtering. Therefore, the face area can be restored with high resolution. Regions other than the face region (background, human body) are encoded after filtering. Here, movement in the fine portion, noise, etc. are removed by filtering in the time axis direction, and fine patterns etc. are removed by filtering in the spatial direction. Therefore, the area other than the face is restored with low resolution. As a result, the face area is displayed clearly as the restored entire image, and the area other than the face (background, etc.) is displayed in a blurred manner.

However, when the face area is detected, if there is almost no change between the previous frame image and the current frame image, the face area is not detected. In this case, the entire input image is directly encoded without any filtering. (Because there are certain undetectable face areas present in the whole image) Therefore the whole restored image is displayed sharply.

When a person (presenter or the like) is input as an image, the facial expression may or may not change. Therefore, when this image is encoded and transmitted and restored, the face area is always displayed clearly, but the background portion is clear or blurred, and the resolution thereof changes from moment to moment. That is, there is a drawback in that it is difficult for the viewer of the restored image to see it due to the change in the display mode of the background portion that is originally still.

[0006]

As described above, in the conventional image coding apparatus, the background portion, which should be originally stationary, is clearly displayed or blurred, and therefore has a drawback that it is very difficult to see. There was An object of the present invention is to provide an image coding apparatus in which the display mode of a stationary background or the like does not change.

[0007]

According to the present invention, a specific area such as a face area is detected in an input image signal and this specific area designating signal is output. It also detects a moving area (including a specific area) in the input image signal and outputs this moving area designating signal. According to the specific area designating signal and the moving area designating signal, the image signals of the static areas other than the moving area in the input image signal are in the time axis direction, and the image signals of the moving area other than the specific area are in the time axis direction and the space. Filter in the direction. The filtered output image signal is encoded by the encoding circuit.

[0008]

When the image signal of the moving image is reduced by the low-pass filter in the time axis direction and the space direction before the image signal of the moving image is reduced, the time axis direction and the space are reduced like the inter-frame DCT encoding. When moving picture coding is performed using the correlation of directions, the amount of information generated by quantization is reduced.

In the present invention, for example, in the case of an image signal of a human image, filtering processing in the time axis direction is performed on a still area other than the moving area, and low filtering in the time axis direction and the spatial direction is performed on a moving area other than the face area. By performing the band-pass filter process, the amount of information generated in regions other than the face region is reduced during quantization, and a stationary background is not blurred.

[0010]

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

FIG. 1 is a block diagram of an image coding apparatus according to an embodiment of the present invention. First, the signal of the entire image of the current frame is input from the input terminal 1 to the face area detection circuit 4. The face area detection circuit 4 stores the entire image of the previous frame, and compares this with the entire image of the current frame in pixel units. The current pixel whose grayscale value changes compared to the previous pixel at the same position is picked up, and these are set as a moving area (for example, a person). Next, the face area is detected by examining the area of the moving area (for example, the elliptical area located above the person is the face). Further, for the detected face area, the circumscribed rectangle is obtained and the end point address is stored (this is the official face area). Next, the main processing of image coding is started. First, an image signal for each pixel of the entire image of the current frame is input to the time axis filter 2, the face area detection circuit 4, and the moving area detection circuit 5 via the input terminal 1. Similar to the face area detection circuit 4, the moving area detection circuit 5 stores the entire image of the previous frame, and compares the gray value of the input pixel with the gray value of the previous frame pixel at the same position. If it is greatly changed, it is regarded as a moving area and a moving area designating signal is output. The face area detection circuit 4 also outputs a face area designation signal when the address of the input pixel is within the stored address range of the face area. When the face area designation signal is output (ON), it is inverted by the inverter. Therefore, the time axis filter 2 is turned off. On the contrary, when the face area designation signal is not output (OFF), the time axis filter 2 is turned ON.
Becomes Further, when the motion area designating signal is output and the face area designating signal is not output (for example, the body of a person), the ON signal is output from the AND circuit, so that the spatial filter 3 turns off.
Become N. When both the moving area designating signal and the face area designating signal are output (a person's face portion), the AND circuit outputs OF.
Since the F signal is output, the spatial filter 3 is turned off. Therefore, when the input image signal is an image of a person, only the areas other than the face area (background, body of the person)
Are filtered by the time axis filter 2. In addition, only the moving area (human body part) other than the face area is spatially filtered 3
Filtered by. The time axis filter 2 removes minute pattern changes (for example, the pattern of a person's clothes is displaced due to body movement) and noise. The spatial filter 3 removes the fine pattern itself (the pattern itself of a person's clothes, etc.). Therefore, in the case of an image signal showing a person, the time axis filter 2 always cuts noise and minute movements of the body and background of the person. Further, with respect to the body of the person, a finer pattern is cut by the spatial filter 3. The face area is not filtered. In other words, since the background portion is always subjected to constant filtering, the background portion of the restored image has a constant display state for the viewer. Motion area (face area of person,
Since there may be a case where the body part cannot be detected when there is no change from the previous frame, the display mode may change depending on the presence or absence of filtering. However, since the moving area (person) is an object that originally moves, the change does not bother the viewer. The image signal selectively filtered by the time axis filter 2 and the spatial filter 3 is input to the encoder 6, encoded, and output from the output terminal 7. The encoder may be any encoder as long as it compresses information by utilizing the correlation between the time axis direction and the spatial direction. For example, international standard H.264. 261 may be used. FIG. 2 is a diagram showing an example of area division when one person is shown in the center of the screen. Area 1 is the face area,
Area 2 is a static area (background) other than the face area, and area 3 is a moving area (body part) other than the face area. As described above, FIG.
In the example, region 1 is not filtered,
The area 2 is a time axis filter only, and the area 3 is filtered in the time axis direction and the space direction. The time axis filter is applied to the area 2 mainly to suppress the noise component, but if the noise is small, it is not necessary to filter the area 2 either. Therefore, like the spatial filter, the time axis filter may be turned on only in the region 3.

FIG. 3 is a block diagram showing an example of the time axis filter of FIG. The image input from the input terminal 10 is subtracted in pixel units from the image one frame before from the frame memory 15 by the subtracter 12, and the nonlinear circuit 13
Entered in. And the filter is turned on from the input terminal 11.
The / OFF signal is also input. The signal non-linearly converted by the non-linear circuit 13 and the input image signal are added by the adder 14 and output from the output terminal 16. The output result is stored in the frame memory 15 as the current frame image. Figure 4
An example of the characteristics of the non-linear circuit 13 is shown in FIG. (A) is when the filter is off, and (b) is when the filter is on. In the case of (a), the small-amplitude part (noise etc.) is strongly filtered in the time axis direction, and in the case of (b), it is strong in the small-amplitude part and weak in the large-amplitude part (fine pattern movement). Axial filter is applied. That is, when the filter is OFF, it functions as a noise reducer (noise suppression), and when the filter is ON, it also serves as a noise reducer and a time axis filter.

FIG. 5 is a block diagram showing an example of the spatial filter of FIG. In FIG. 5, the pixel memory 22 and the line memory 23 delay the output signal of this filter by one pixel and one line, respectively. The outputs of the memories 22 and 23 are added by the adder 24 and are halved by bit shifting. The input image signal is subtracted from the signal by the subtractor 25 and input to the coefficient multiplier 26. Further, a filter ON / OFF signal is also input from the input terminal 21. The coefficient multiplier outputs 0 when the filter is OFF, and the coefficient α (0 <α <when the filter is ON.
The value multiplied by 1) is output. The signal and the input image signal are added by the adder 27 and output from the output terminal 28.
By this spatial filter, when the grayscale value of the input pixel is largely changed as compared with the grayscale values of the adjacent pixels (one pixel before and one line before), the grayscale value is changed to a smaller grayscale value. In other words, the fine pattern itself (clothes pattern, etc.) is blurred.

FIG. 6 is a block diagram showing an example of the moving area detection circuit 5 of FIG. The image signal of the current frame is input to the input terminal 30, and the image signal of the previous frame is input to the input terminal 31. If the image signal of the previous frame is brought from the output of the frame memory in FIG. 3, the frame memory can be saved. Then, the subtracter 32 calculates the inter-frame difference on a pixel-by-pixel basis, the absolute value circuit 33 calculates the absolute value, and the comparator 34 compares the absolute value with the threshold value TH to set a portion larger than the threshold value as a moving area. The face area detection circuit 4 is realized by providing a circuit for detecting only the face area from the area, position, and shape of the moving area in the subsequent stage of the configuration of FIG.

[0015]

According to the present invention, for example, when an image signal of a human image is targeted, a region other than the face region (background, body) is subjected to filter processing in the time axis direction, and a moving region (body) other than the face region is processed. By performing the filtering process in the time axis direction and the space direction on (), the still area such as the background is always displayed in a fixed manner in the restored image. Therefore, for the viewer, the restored image can be viewed in a state close to natural.

[Brief description of drawings]

FIG. 1 is a block diagram of a pixel encoding device according to an embodiment of the present invention.

FIG. 2 is a diagram showing an example of region division.

FIG. 3 is a block diagram showing an example of the time axis filter of FIG.

FIG. 4 is a block diagram showing an example of characteristics of the nonlinear circuit of FIG.

5 is a block diagram showing an example of the spatial filter of FIG.

FIG. 6 is a block diagram showing an example of a moving area detection circuit in FIG.

[Explanation of symbols]

 DESCRIPTION OF SYMBOLS 1 ... Input terminal 2 ... Time axis filter 3 ... Spatial filter 4 ... Face area detection circuit 5 ... Motion area detection 6 ... Encoder 7 ... Output terminal

Claims (1)

[Claims] 1. An image coding apparatus for coding an input image signal of a moving image, a moving area detecting means for detecting a moving area in the input image signal, and outputting the moving area designating signal, the input image. Detecting a specific area included in the moving area in the signal, a specific area designating means for outputting this specific area designating signal, according to the moving area designating signal and the specific area designating signal, other than the moving area in the input image signal Filtering means for filtering the area in the time axis direction and filtering in the time axis direction and the spatial direction for areas other than the specific area in the moving area in the input image signal, and the image signal filtered by this filter means An image coding apparatus comprising: a coding unit for coding.