JPH09107543A - Quantizer control system - Google Patents

Abstract

PROBLEM TO BE SOLVED: To attain efficient quantization by comparing a generated information amount between inter-frame coding units with a prescribed threshold level so as to discriminate whether the image is in a still image or in a moving area, and distinguishing the still area image into a still image with a fine pattern and a mere still image. SOLUTION: A changeover detection circuit 36 detects channel chageover and gives the result to a coding control section 37 to reset quantization of a coding section 34 and to allow the coding section 34 to retry coding control. Furthermore, the coding section 34 provides image data in each stage of coding to a coding control section 37 and a transmission buffer 35 gives a residual amount signal of it to a coding control section 37. The coding control section 37 compares the generated information amount between different inter-frame coding units with a prescribed threshold level to discriminate a still area or a moving area and the coding section 34 applies rough preset quantization to the still area and applies fine quantization to the moving area. Furthermore, the control section 37 discriminates a fine pattern based on the generated amount of higher order DCT coefficients for the still area to control the coding section 34. Thus, efficient quantization is attained.

Description

Detailed Description of the Invention

[0001]

BACKGROUND OF THE INVENTION 1. Field of the Invention The present invention relates to a quantizer control system in moving picture coding. A technique of compressing and encoding image data has been used conventionally. In recent years, a video conference system has been used, and an efficient encoding technique for image data has been demanded.

[0002]

2. Description of the Related Art When encoding a moving image, the coarseness of quantization,
The dense step (quantization step) is closely related to the image information amount and the image quality. FIG. 9 is a diagram showing the relationship between the quantization step and the generated information amount. The vertical axis represents the generated information amount, and the horizontal axis represents the quantization step. When the quantization step becomes dense,
The image quality is improved. Low transmission rate (64 Kb / s, 128
In Kb / s, etc., the amount of information that can be transmitted is limited. Therefore, when the amount of generated information increases, it is necessary to reduce the image quality and reduce the amount of data.

When performing quantization, for example, as shown in FIG. 10, an image is divided into a plurality of (here, 12) blocks (coding units), and the remaining amount in the transmission buffer is monitored, Quantization step (number of quantization bits)
Is determined. For example, when the remaining amount of the transmission buffer is small, coarse quantization is performed, and when the remaining amount of the transmission buffer is large, dense quantization is performed. FIG. 11 is a diagram showing the relationship between the quantization step and the transmission buffer. The vertical axis represents the quantization step, and the horizontal axis represents the remaining amount of the transmission buffer. It can be seen that when the remaining amount of the transmission buffer becomes large, the quantization step becomes dense, and when the remaining amount of the transmission buffer becomes small, the quantization step becomes coarse.

In video transmission of a low bit rate (for example, 64 Kbps) in a video conference system or a video telephone system, it is necessary to efficiently use a small transmission capacity. Therefore, the interframe predictive coding method is used.
In the inter-frame predictive coding method, a difference between image data between two consecutive frames is extracted, and only the difference value is encoded and transmitted. In addition, there is a method in which a static / moving area is determined based on this difference value for each encoding unit and encoding is performed.

FIG. 12 is an explanatory diagram of a still area / moving area determination method in a certain coding unit. In the figure, FL1 to
FL3 is a frame. These frames are divided into a predetermined size to form a coding unit U. The difference of the image data between frames is obtained for each unit corresponding to the encoding unit U. For example, the difference of the image data between the coding units U of the frames FL1 and FL2 is calculated, and the difference of the image data between the coding units U of the frames FL2 and FL3 is calculated. If the difference thus obtained is large, the coding unit is determined to be a moving area, and if it is small, it is determined to be a still area.

[0006]

In the above-described interframe predictive coding method, since the amount of information generated is small in the still area, it is necessary to efficiently code the moving area in order to use the transmission capacity efficiently. There is. However, in a general scene such as a video conference system, the upper half of the person is displayed in the center of the screen, and other backgrounds are mostly stationary (see FIG. 10). In such a scene,
As a result of the decrease in the transmission buffer capacity, that is, the increase in the transmission buffer remaining amount in the coding unit of the still portion, fine quantization is performed. That is, a large amount of data is allocated to the still area in one screen.

On the other hand, in order to prevent the accumulation of calculation errors, it is necessary to perform intraframe predictive coding at regular intervals. Intra-frame predictive coding is a method of detecting a change in pixel value between pixels in one frame and performing periodic refreshing. At this time, similarly to the case of the inter-frame predictive coding method, even in the intra-frame predictive coding method, if coarse quantization is performed on the still region, information is generated even in the still image in the intra-frame predictive coding method. The fine pattern in the still area is blurred. For example, the case in which the person's clothes shown in FIG. 10 are clothes with a fine pattern corresponds to this.

Therefore, it is necessary to perform the quantization control by giving more importance to the fineness of the picture than the movement of the image at the time of the periodic refresh, but conventionally, the interframe predictive coding / intraframe predictive coding is not distinguished. Since the moving area / still area is quantized, the fine image in the still area is deteriorated and the overall image quality is deteriorated. Further, if the quantization at the time of intra-frame prediction is simply fined, the amount of generated data becomes excessive, which adversely affects the encoding after the periodic refresh.

The present invention has been made in view of the above problems, and distinguishes a still region with a still image with a thin pattern from a simple still image, and performs efficient quantization with little deterioration in image quality. It is an object of the present invention to provide a quantizer control system capable of performing the above.

[0010]

FIG. 1 is a block diagram showing the principle of the present invention. In the figure, 10 is an encoding means for inputting an image and encoding in an encoding unit, and 20 is an output of the encoding means 10 at a predetermined encoding stage, and is provided between encoding units between different frames. The amount of generated information of is compared with a predetermined threshold to determine whether it is a static area or a moving area,
It is an encoding control means for performing control for predetermined encoding.

According to the structure of the present invention, as a result of the coding control means 20 comparing the amount of information generated between coding units of different frames with a predetermined threshold value to judge whether it is a still area or a moving area, The encoding means 10 can perform encoding according to the result of this determination, and can perform efficient quantization with little deterioration in image quality.

In this case, the encoding control means 2
0 is characterized in that, when the output of the encoding means 10 is received and the corresponding encoding unit is determined to be a still region, coarse quantization set in advance is performed.

According to the structure of the present invention, a small transmission capacity can be efficiently used by performing coarse quantization in the static region. Also, the encoding control means 2
0 is characterized by receiving the output of the encoding means 10 and judging the fineness of the picture in the still area, and finely quantizing the coding unit having the fine picture in the still area.

According to the structure of the present invention, the still area is distinguished from a still picture with a thin picture and a simple still picture, and fine quantization is performed in the still area where a fine picture exists to prevent deterioration of image quality. be able to.

Further, the method for judging the fineness of the pattern is to compare the generation amount of the higher-order coefficient of the DCT coefficient in the coding unit of the intraframe coding in the periodic refresh with a predetermined threshold value. Is characterized by.

According to the structure of the present invention, the fineness of the pattern can be efficiently judged by using the method of comparing the generation amount of the higher-order coefficient of the DCT coefficient with the predetermined threshold value. The predetermined threshold value is changed according to the transmission rate.

According to the structure of the present invention, by changing the predetermined threshold value according to the transmission rate, it is possible to perform efficient quantization according to the transmission rate. Further, the encoding unit for judging the fineness of the pattern is a macro block.

According to the structure of the present invention, efficient quantization can be performed by using a macroblock as a coding unit. The encoding unit for judging the fineness of the design is GOB.

According to the structure of the present invention, by using GOB as a coding unit, efficient quantization can be performed. In addition, it is characterized in that the predetermined thresholds are different depending on the encoding frame format of CIF and QCIF.

According to the configuration of the present invention, by varying the predetermined threshold value according to the frame format, it is possible to perform optimum quantization according to the type of the frame format.

[0021] Further, the present invention is characterized in that the input image is obtained from a TV camera, and in a TV conference system having a plurality of TV cameras, the channels of these TV cameras are switched and used.

According to the structure of the present invention, the desired image can be quantized by switching the channels of the television camera. Further, the encoding control means 20 is provided with a channel switching detection means for detecting the channel switching, and when the coding control means 20 receives the output of the channel switching detection means and detects that the channel has been switched, the coding means 10 performs quantization. It is characterized by resetting and controlling.

According to the structure of the present invention, when the channel is switched, the optimum quantization can be performed according to the type of the channel. Further, since the target of each channel is different, when the quantizer control is different for each channel, the encoding control means 20 changes the quantization control when detecting the channel switching.

According to the structure of the present invention, the optimum quantization can be performed according to the type of channel. Further, the method for judging the fineness of the pattern is characterized by comparing the generation amount of the higher-order coefficient of the DCT coefficient in the coding unit of the intraframe coding in the periodic refresh with a predetermined threshold value. There is.

According to the structure of the present invention, the fineness of the pattern can be efficiently determined by using the method of comparing the generation amount of the higher-order coefficient of the DCT coefficient with the predetermined threshold value. Further, the method for judging the fineness of the pattern is to compare the generation amount of the higher-order coefficient of the DCT coefficient in the coding unit of the intraframe coding in the periodic refresh with a predetermined threshold value, The feature is that it is changed according to the transmission rate.

According to the structure of the present invention, by using the method of comparing the generation amount of the higher-order coefficient of the DCT coefficient with the predetermined threshold value, the fineness of the pattern can be efficiently judged and the transmission rate can be improved. By changing the predetermined threshold value accordingly, efficient quantization can be performed according to the transmission rate.

Further, the method for judging the fineness of the pattern is to compare the generation amount of the higher-order coefficient of the DCT coefficient in the coding unit of the intraframe coding in the periodic refresh with a predetermined threshold, It is characterized in that the coding unit for judging the fineness of the pattern is a macro block.

According to the structure of the present invention, by using the method of comparing the generation amount of the higher-order coefficient of the DCT coefficient with the predetermined threshold value, the fineness of the pattern can be efficiently judged, and the coding unit can be improved. By using a macroblock as, efficient quantization can be performed.

Further, the method for judging the fineness of the pattern is to compare the generation amount of the higher-order coefficient of the DCT coefficient in the coding unit of the intraframe coding in the periodic refresh with a predetermined threshold value. The encoding unit for judging the fineness of the pattern is GOB.

According to the structure of the present invention, by using the method of comparing the generation amount of the higher-order coefficient of the DCT coefficient with the predetermined threshold value, the fineness of the pattern can be efficiently judged and the coding unit can be improved. By using GOB as, it is possible to perform efficient quantization.

Further, the method for judging the fineness of the pattern is to compare the amount of higher-order coefficients of DCT coefficients in the coding unit of intraframe coding in the periodic refresh with a predetermined threshold value. It is characterized in that the threshold value is changed according to the transmission rate.

According to the structure of the present invention, by using the method of comparing the generation amount of the higher-order coefficient of the DCT coefficient with the predetermined threshold value, the fineness of the pattern can be efficiently judged and the transmission rate can be improved. Accordingly, efficient quantization can be performed.

Further, the method for judging the fineness of the pattern is to compare the amount of higher-order coefficients of DCT coefficients in the coding unit of intraframe coding in the periodic refresh with a predetermined threshold value. It is characterized in that the coding unit for judging the fineness of the pattern is a macro block.

According to the structure of the present invention, by using the method of comparing the generation amount of the higher-order coefficient of the DCT coefficient with the predetermined threshold value, the fineness of the pattern can be efficiently judged, and the coding unit can be improved. By using a macroblock as, efficient quantization can be performed.

Further, the method for judging the fineness of the pattern is to compare the amount of higher-order coefficients of DCT coefficients in the coding unit of intraframe coding in the periodic refresh with a predetermined threshold value. The encoding unit for judging the fineness of the pattern is GOB.

According to the structure of the present invention, by using the method of comparing the generation amount of the higher-order coefficient of the DCT coefficient with a predetermined threshold value, the fineness of the pattern can be efficiently judged and the coding unit can be improved. By using GOB as
It is possible to perform efficient quantization.

[0037]

Embodiments of the present invention will be described below in detail with reference to the drawings. 2 is a block diagram showing an embodiment of the present invention, and FIG. 3 is a conceptual diagram of a video conference system which is a premise of the block diagram shown in FIG. First, the video conference system of FIG. 3 will be described. In this video conference system, a plurality of conferees 2 (to 5 people) are sitting on a semicircular table 1. These conference participants are photographed by a plurality of cameras 3 and 4 arranged in front of them. 3
Is a fixed camera that mainly captures the entire conference. On the other hand, 4 is a moving camera that rotates and moves in the direction shown in the figure, and is used to capture the speaker of the conference.

The images taken by the fixed camera 3 and the moving camera 4 are subjected to predetermined image compression and transmitted to the host side, and the compressed image sent from the host side is decompressed and displayed on the large television monitor 5. It is supposed to be displayed. It should be noted that although an example in which two cameras are installed is shown in the figure, there are the following types of cameras used in the video conference system. Type with one camera with swivel and camera moving Type with one camera fixed with two cameras Type with two swivel with both cameras moving Two cameras with one fixed, The rest is of the type that has N input channels and is equipped with an optional camera or the like. Here, the type of is used, but as shown in May be used. Further, the captured images of the fixed camera 3 and the moving camera 4 are transmitted not only when they are combined, but also when the respective captured images are switched by a switch and transmitted to the host side.

Next, FIG. 2 will be described. The same parts as those in FIG. 3 are designated by the same reference numerals. In the figure, 30
Is a camera movement control unit that controls movement of the moving camera 4 within a predetermined range. Reference numeral 31 is a channel switching unit that receives the outputs of the fixed camera 3 and the moving camera 4 and switches the channel by a switching signal, 32 is an A / D conversion unit that converts an analog image signal of an incoming channel into digital image data, 33 Is a format conversion unit that receives the output of the A / D conversion unit 32 and converts the image format. Television image formats include NTSC (525 × 60 system) used in Japan and PAL (625 × 50 system) used in Europe and the like, and the format conversion unit 33 converts the intermediate format CIF between these formats. (352 horizontal pixels x 288 vertical pixels
Pixels) are used for mutual conversion.

Reference numeral 34 is an encoding unit for compressing and encoding the image data having passed through the format conversion unit 33 by a predetermined method. Reference numeral 35 is a transmission buffer for temporarily storing the encoded data, the output of which is a transmission line. Go out to. Reference numeral 36 denotes a switching detection circuit that receives the output of the channel switching unit 31 to detect the time of channel switching, and 37 receives the outputs of the switching detection circuit 36, the encoding unit 34, and the transmission buffer 35, and the mutual encoding units between different frames It is an encoding control unit that compares the amount of information generated during a period with a predetermined threshold to determine whether the region is a still region or a moving region, and also performs operations for various encoding controls. In the above description, the channel switching unit 31, the A / D conversion unit 32, the format conversion unit 33, the coding unit 34, and the transmission buffer 35 constitute the coding unit 10 in FIG. 1, and the switching detection circuit 36 and the coding unit The control unit 37 constitutes the encoding control means 20 shown in FIG.

Reference numeral 38 is a reception buffer for receiving and temporarily storing image data sent from the host side from the host side.
Reference numeral 9 denotes a decoding unit that receives the output of the reception buffer 38 and decodes image data, 40 is a format conversion unit that converts the image format as necessary, and 41 is an analog that receives the output of the format conversion unit 40. It is a D / A conversion unit for converting into an image signal. The output of the D / A converter 41 is connected to the large television monitor 5. The operation of the system configured as described above will be described below.

The image signal photographed by the fixed camera 3 or the moving camera 4 enters the channel switching unit 31, and the image signal selected by the channel switching unit 31 continues A / D.
The conversion unit 32 converts the digital image data.
The image data converted into digital data enters the subsequent format conversion unit 33 and is converted into a predetermined format. Then, the output of the format conversion unit 33 enters the encoding unit 34 and is subjected to predetermined encoding. The encoded image data enters the transmission buffer 35 and is transmitted from the transmission buffer 35 to the host side via the transmission path.

On the other hand, the image data sent from the host side via the transmission path is received by the reception buffer 38,
It is temporarily stored. The image data stored in the reception buffer 38 is decoded by the subsequent decoding unit 39. The output of the decoding unit 39 enters the format conversion unit 40 and is converted into a required format. The format-converted image data is converted into an analog image signal by the subsequent D / A converter 41, and then sent to the television monitor 5 for display.

Next, details of the encoding operation of the system will be described. The switching detection circuit 36 detects the channel switching and supplies it to the coding control unit 37, the coding unit 34 supplies the image data at each stage of coding to the coding control unit 37, and the transmission buffer 35 stores the buffer data. The remaining amount signal is given to the encoding control unit 37. Here, the encoding control unit 3
7 receives these signals and gives a predetermined encoding control command to the encoding unit 34 to cause the encoding unit 34 to perform encoding.

In this case, the coding control section 37 can compare the amount of information generated between coding units between different frames with a predetermined threshold value to determine whether it is a still area or a moving area (interframe). Predictive coding). Since the encoding unit 34 can perform encoding according to the determination result, efficient quantization can be performed without degrading image quality.

As a result of the above judgment, when the corresponding coding unit is judged to be the still area, the coding unit 34 performs the coarse quantization set in advance, and the corresponding coding unit is judged to be the moving area. If so, dense quantization is performed. As a result, a small transmission capacity can be used efficiently.

Note that it is necessary to switch channels in a video conference system in which images are taken by a plurality of cameras. For example, in a 2-channel video conference system,
One camera has a swivel base and is movable, and the moving camera 4 also moves each time the speaker changes for the purpose of focusing on the conference speaker who speaks, and mainly one person is enlarged. However, since the other fixed camera 3 is equipped with a camera intended to capture the entire conference room, there is a difference in the image object between channel 1 and channel 2.

Therefore, by switching the channel, the image of the desired channel can be quantized.
When channel switching occurs, images before and after switching do not have continuity, so that determination of a still area and a moving area using interframe predictive coding before switching becomes meaningless. The video at the moment of switching is determined to be a moving region by the encoding control unit 37 and is densely quantized. However, due to human visual characteristics, it is not possible to follow an instantaneous change, so that the amount of data generated increases. Has little effect on improving image quality.

Therefore, when the switching detection circuit 36 detects the channel switching, it notifies the coding control unit 37 of that. Upon recognizing the channel switching, the coding control unit 37 resets the quantization of the coding unit 34 and performs the coding control again. This makes it possible to perform optimum quantization according to the type of channel.

When the coding control unit 37 detects that the channel has been switched, the target imaged in each channel is different, so that the coding control unit 37 gives a control signal to change the quantization control. . This makes it possible to perform optimum quantization according to the type of channel.

Further, in the present invention, the coding control unit 37 determines the still area and the moving area by the above-mentioned interframe predictive coding method (see FIG. 12), receives the output of the coding unit 34, and receives the still picture. The fineness of the pattern in the case of the area is determined, and the encoding unit 34 is controlled based on the result. As a result, the still region is distinguished from a still image with many patterns and a simple still image, and even in the still region, fine quantization is performed in a region where a fine pattern exists, and in a simple still image, coarse quantization is performed to deteriorate image quality. Can be prevented.

Next, how to determine the fineness of the pattern of a still image will be described. FIG. 4 is an explanatory diagram of an example of encoding and decoding according to the present invention. As shown in the figure, it is assumed that there is CIF format image data of 288 vertical pixels × 352 horizontal pixels. This one image is divided into 8 × 8 blocks, and a block of 8 × 8 pixels is used as a coding unit. The image of 8 × 8 pixels is as shown in (a), and there are 64 pixels a1 to a64. DCT conversion is performed on the image data in this encoding unit. DCT conversion (Discrete Cosine T)
(Transform: Discrete Cosine Transform) converts image data into a frequency signal. As a result, (b)
The converted image as shown in is obtained. The converted image data consists of 64 pieces of data from b1 to b64. The upper left part is the low frequency term, and the lower right part is the high frequency term.

The image data after the DCT conversion is divided using the reference quantization table. Here, a case where the value is divided by a certain value Q is shown. After division and rounding the remainder, quantized data as shown in (c) is obtained. The quantized data is biased to the low frequency term. Incidentally, c1 in (c) is a DC component. The encoded data thus obtained is transmitted to the host side via the transmission line.

When the transmitted quantized data is inversely quantized, data as shown in (d) is obtained. Pixel values are low frequency terms only. When this is DCT inverse transformed, (e)
The image data as shown in is reproduced. Here, (e)
A'is almost equal to a in (a), which means that the image is reproduced.

In the present invention, the fineness of the pattern in the still image is judged from the data shown in (b) after the DCT conversion. FIG. 5 is an explanatory diagram of an algorithm for determining the fineness of the pattern, and shows the operation of the encoding control unit 37. The encoding control unit 37 inputs the data after the DCT conversion executed by the encoding unit 34 and makes the following determination.

When DCT conversion is performed on the image data of 8 × 8 pixels (not necessarily 8 × 8) shown in (a), frequency data as shown in (b) is obtained. The upper left part is the low frequency term and the lower right part is the high frequency term. Here, when arranging the data from the low frequency term to the high frequency term,
(C). Here, H is a header added to the beginning of the data. This header indicates the beginning of a unit. Here, the range 1 which is the criterion of the determination is determined from the result of the image quality investigation using several kinds of representative images, and the coefficient (for example, b60 to b64) occurring after the range 1 is determined by a certain threshold value (experimental determination) When it is larger than that of (1), it is determined that the pattern is fine. When the coefficient (b60 to b64) after the range 1 is smaller than a certain threshold value, it is determined that the pattern is normal. By using such an algorithm, the fineness of the pattern can be efficiently determined.

In this way, the encoding control unit 37 determines the fineness of the pattern of the still image, and the encoding unit 34 is determined based on the result.
Control. That is, even in the case of a still image, when the pattern is fine, dense quantization is performed. It should be noted that coarse quantization is performed for normal still images. As a result, it is possible to perform quantization that can efficiently use the transmission capacity.

In this case, the encoding controller 37
The threshold is made variable according to the transmission rate of the transmission line. Thereby, efficient quantization can be performed according to the transmission rate.

FIG. 6 shows a switch detection circuit 36 (see FIG. 2).
It is a figure which shows the example of a specific structure. 36a and 36b are flip-flops, and 36c is an exclusive OR gate (EOR gate). The first flip-flop 36a receives the channel switching signal output from the channel switching unit 31 and the frame pulse, and the second flip-flop 36b receives the output of the first flip-flop 36a and the frame pulse. . Exclusive OR gate 36c
Includes first and second flip-flops 36a and 36b.
Contains the output of each. Then, a switching detection signal is output from the exclusive OR gate 36c and enters the encoding control unit 37. One frame pulse per frame
The channel switching signal is generated asynchronously with the frame pulse. The operation of the circuit thus configured will be described as follows.

First, when the channel switching signal "1" enters the flip-flop 36a, this channel switching signal is latched at the first frame pulse. Next, 2
When the th frame pulse is input, the output "1" of the first flip-flop 36a changes to the second flip-flop 36b.
No channel switching (“0”) is latched in the first flip-flop 36a. Since the outputs of the first and second flip-flops 36a and 36b are input to the exclusive OR gate 36c, when the channel is switched, these two inputs become different, and the exclusive OR gate 36c The sum gate 36c outputs "1". This "1" level indicates that there has been channel switching (change). On the other hand, when there is no channel switching, the first and second flip-flops 36a, 3a
Since the outputs of 6b both have the same value, the output of the exclusive OR gate 36c becomes "0". That is, when the output of the exclusive OR gate 36c is "1", there is channel switching, and when the output of the exclusive OR gate 36c is "0", there is no channel switching.

In FIG. 7, the coding unit is macroblock (M
It is explanatory drawing of the macroblock when it is set to B). The macroblock is 16 pixels (pixels) x 16 lines of Y
It is composed of a (luminance) signal and color difference signals U and V of 8 pixels × 8 lines located in the middle. This macroblock is composed of 8 × 8 blocks. Each block consists of four Y data and one U / V data. The luminance signal Y and the color difference signal U / V are R, G, and B.
The signals are represented by the following equations with ER, EG, and EB respectively.

Y = 0.3 ER +0.59 EG +0.11 EB (1) U = ER −Y (2) V = EB −Y (3) As described above, by using the macro block as the coding unit, the efficiency is improved. Good quantization can be performed.

The unit of coding is not limited to macroblocks, but other types of units can be used. For example, GOB (Group Of macroBloc
k) can be used. FIG. 8 is an explanatory diagram of GOB. (A) is the CIF format, (b) is the QCI
F (1/4 of CIF) is shown. ITU-TH. 261
Defines the CIF and QCIF formats as encoded frames. GOB corresponds to 1/12 of CIF and 1/3 of QCIF, and Y signal is 176 pixels × 48
It is composed of color difference signals U and V of 88 pixels × 24 lines which spatially correspond to each line. Further, GOB is divided into 33 macroblocks (MB) as shown in (c).

Thus, by using GOB as a coding unit, efficient quantization can be performed. Also, the coding frame formats are CIF and Q.
The CIF and the threshold in the above-described algorithm for determining the fineness of the pattern can be made different. This makes it possible to perform optimum quantization according to the type of frame format.

[0065]

As described above in detail, according to the present invention, an encoding means for receiving an input image and performing encoding in a certain encoding unit, and a predetermined encoding step of the encoding means Receiving the output, the amount of information generated between coding units of different frames is compared with a predetermined threshold value to determine whether it is a still area or a moving area, and also to perform control for predetermined coding. By including the control means, the coding control means compares the amount of information generated between coding units between different frames with a predetermined threshold value to determine whether the area is a still area or a moving area. Can perform encoding according to the result of this determination, and can perform efficient quantization with little deterioration in image quality.

In this case, the coding control means receives the output of the coding means, and when the corresponding coding unit is determined to be the still area, performs the coarse quantization set in advance. As a result, it is possible to efficiently use a small transmission capacity by performing coarse quantization in the static region.

Further, the encoding control means receives the output of the encoding means, judges the fineness of the picture in the still area, and finely quantizes the coding unit in the still area with the fine picture. By doing so, it is possible to prevent the deterioration of the image quality by distinguishing the still region from a still image with many patterns and a simple still image, and performing fine quantization in the still region where a fine pattern exists.

Further, the method for judging the fineness of the pattern is to compare the generation amount of the high-order coefficient of the DCT coefficient in the coding unit of the intraframe coding in the periodic refresh with a predetermined threshold value. Thus, the fineness of the design can be efficiently determined.

By changing the predetermined threshold value according to the transmission rate, the predetermined threshold value is changed according to the transmission rate, so that efficient quantization can be performed according to the transmission rate.

Further, since the coding unit for judging the fineness of the pattern is a macroblock, efficient quantization can be performed by using the macroblock as a coding unit.

Further, since the coding unit for judging the fineness of the pattern is GOB, GO is used as the coding unit.
By using B, efficient quantization can be performed.

The encoding frame format is CI
By making the predetermined threshold different between F and QCIF, and making the predetermined threshold different depending on the frame format, it is possible to perform optimum quantization according to the type of the frame format.

Further, in the video conference system in which the input image is obtained from the television camera, and the channels of the television cameras are switched and used in the television conference system, the channels of the television camera are switched. Quantization of the desired image can be performed.

Further, it comprises a channel switching detecting means for detecting the channel switching, and when the coding control means receives the output of the channel switching detecting means and detects that the channel has been switched, the coding means quantizes it. By resetting and performing control, when the channel is switched, optimum quantization can be performed according to the type of channel.

Further, since the target of each channel is different, when the quantizer control is different for each channel, when the coding control means detects channel switching, the coding control is changed to change the channel type. It is possible to perform optimum quantization in accordance with the above.

Further, the method of judging the fineness of the pattern is to compare the generation amount of the higher order coefficient of the DCT coefficient in the coding unit of the intraframe coding in the periodic refresh with a predetermined threshold value. Thus, the fineness of the design can be efficiently determined.

Further, the method for judging the fineness of the picture pattern is to compare the generation amount of the higher-order coefficient of the DCT coefficient in the coding unit of the intraframe coding in the periodic refresh with a predetermined threshold value. By changing the threshold according to the transmission rate, it is possible to efficiently determine the fineness of the pattern, and by changing the predetermined threshold according to the transmission rate, efficient quantization according to the transmission rate can be achieved. Can be done.

Further, the method of judging the fineness of the pattern is to compare the generation amount of the higher-order coefficient of the DCT coefficient in the coding unit of the intraframe coding in the periodic refresh with a predetermined threshold value. Since the encoding unit for determining the fineness of the pattern is the macroblock, the fineness of the pattern can be efficiently determined, and the efficient quantization can be performed by using the macroblock as the encoding unit. Can be done.

Further, the method of judging the fineness of the pattern is to compare the generation amount of the higher-order coefficient of the DCT coefficient in the coding unit of the intraframe coding in the periodic refresh with a predetermined threshold value. Since the encoding unit for determining the fineness of the pattern is GOB, the fineness of the pattern can be efficiently determined, and G is used as the encoding unit.
By using OB, efficient quantization can be performed.

Further, the method of judging the fineness of the pattern is to compare the generation amount of the higher-order coefficient of the DCT coefficient in the coding unit of the intraframe coding in the periodic refresh with a predetermined threshold value, By changing the threshold value according to the transmission rate, it is possible to efficiently determine the fineness of the pattern and to perform efficient quantization according to the transmission rate. Further, the method of judging the fineness of the pattern is to compare the amount of higher-order coefficients of DCT coefficients in the coding unit of intraframe coding in the periodic refresh with a predetermined threshold value. Since the coding unit that determines the fineness is a macroblock,
It is possible to efficiently determine the fineness of a picture, and by using a macroblock as a coding unit, efficient quantization can be performed.

Further, the method for judging the fineness of the pattern is to compare the amount of higher-order coefficients of DCT coefficients in the coding unit of intraframe coding in the periodic refresh with a predetermined threshold value. Since the encoding unit for determining the fineness of the pattern is GOB, the fineness of the pattern can be efficiently determined, and G is used as the encoding unit.
By using OB, efficient quantization can be performed.

As described above, according to the present invention, the quantizer control system capable of distinguishing the still region between a still image with a thin pattern and a simple still image and performing efficient quantization with little deterioration in image quality. Can be provided. Further, in a system that performs coarse quantization in a still area in moving image transmission, noise (blur) that has occurred can be reduced when the coding unit judged to be the still area has a fine pattern, and the expression is reduced. It becomes appropriate, the amount of information can be effectively allocated, and the overall image quality can be improved.

[Brief description of the drawings]

FIG. 1 is a principle block diagram of the present invention.

FIG. 2 is a block diagram showing an embodiment of the present invention.

FIG. 3 is a conceptual diagram of a video conference system.

FIG. 4 is an explanatory diagram of an example of encoding and decoding.

FIG. 5 is an explanatory diagram of an algorithm for determining the fineness of a pattern.

FIG. 6 is a diagram showing a specific configuration example of a switching detection circuit.

FIG. 7 is an explanatory diagram of macroblocks.

FIG. 8 is an explanatory diagram of GOB.

FIG. 9 is a diagram showing a relationship between a quantization step and an amount of generated information.

FIG. 10 is a diagram showing an example in which an image is divided into 12 coding units.

FIG. 11 is a diagram showing a relationship between a quantization step and a transmission buffer.

FIG. 12 is an explanatory diagram of a still area / moving area determination method in a certain coding unit.

[Explanation of symbols]

 10 Encoding means 20 Encoding control means

Claims (18)

[Claims] 1. An encoding unit for inputting an image and encoding the image in a certain encoding unit, and receiving an output of the encoding unit at a predetermined encoding stage, between encoding units between different frames. A quantizer control system configured to include coding control means for comparing the generated information amount with a predetermined threshold to determine whether the area is a still area or a moving area, and for performing control for predetermined coding. . 2. The encoding control means receives the output of the encoding means and, when the corresponding encoding unit is determined to be a still area, performs a coarse quantization set in advance. A quantizer control system according to claim 1, characterized in that: 3. The encoding control means receives the output of the encoding means and judges the fineness of the pattern in the still area,
3. The quantizer control system according to claim 2, wherein fine quantization is performed for a coding unit having a fine pattern in a still area. 4. The method for judging the fineness of the pattern is to compare the generation amount of higher-order coefficients of DCT coefficients in a coding unit of intraframe coding in periodic refreshing with a predetermined threshold value. 4. The quantizer control system according to claim 3, wherein: 5. The quantizer control system according to claim 4, wherein the predetermined threshold value is changed according to a transmission rate. 6. The quantizer control system according to claim 4, wherein a coding unit for judging the fineness of the pattern is a macroblock. 7. The quantizer control system according to claim 4, wherein the coding unit for judging the fineness of the pattern is GOB. 8. The quantizer control system according to claim 4, wherein the predetermined threshold differs between CIF and QCIF in a coded frame format. 9. The method for obtaining the input image from a television camera, wherein channels of the television cameras are switched and used in a television conference system having a plurality of television cameras. Quantizer control system. 10. A channel switching detection means for detecting the channel switching is provided, wherein the encoding control means receives the output of the channel switching detection means and detects that the channel has been switched,
10. The quantizer control system according to claim 9, wherein the quantizer by the encoding means is reset to perform control. 11. The object of each channel is different, so that when the quantizer control is different for each channel, the coding control means changes the quantization control when detecting channel switching. 10. The quantizer control system described in 10. 12. A method of determining the fineness of the design,
11. The quantizer control system according to claim 10, wherein the generation amount of higher-order coefficients of DCT coefficients in a coding unit of intraframe coding in periodic refresh is compared with a predetermined threshold value. 13. A method for determining the fineness of the design,
The amount of higher-order coefficients of DCT coefficients in a coding unit of intraframe coding in periodic refresh is compared with a predetermined threshold value, and the threshold value is changed according to a transmission rate. 10. The quantizer control system described in 10. 14. A method for determining the fineness of the design,
The amount of higher-order coefficients of DCT coefficients in the coding unit of intraframe coding in the periodic refresh is compared with a predetermined threshold value, and the coding unit for judging the fineness of the pattern is a macroblock. 11. The quantizer control system according to claim 10, wherein: 15. A method for determining the fineness of the design is
The amount of higher-order coefficients of DCT coefficients in the coding unit of the intraframe coding in the periodic refresh is compared with a predetermined threshold, and the coding unit for judging the fineness of the pattern is GOB. 11. The quantizer control system according to claim 10. 16. A method for determining the fineness of the design,
The amount of higher-order coefficients of DCT coefficients in a coding unit of intraframe coding in periodic refresh is compared with a predetermined threshold value, and the threshold value is changed according to a transmission rate. 11. The quantizer control system described in 11. 17. A method for judging the fineness of the design is
The amount of higher-order coefficients of DCT coefficients in the coding unit of intraframe coding in the periodic refresh is compared with a predetermined threshold value, and the coding unit for judging the fineness of the pattern is a macroblock. 12. The quantizer control system according to claim 11, wherein: 18. A method for determining the fineness of the design,
The amount of higher-order coefficients of DCT coefficients in the coding unit of the intraframe coding in the periodic refresh is compared with a predetermined threshold, and the coding unit for judging the fineness of the pattern is GOB. The quantizer control system according to claim 11, wherein: