JP3855332B2 - Image coding method - Google Patents

Description

[0001]
BACKGROUND OF THE INVENTION
The present invention relates to an image encoding method in an image communication apparatus such as a video conference apparatus.
[0002]
[Prior art]
In recent years, ITU-T has officially recommended video coding systems, still picture coding systems, multi-volume systems, communication procedures, etc. for audio-visual services. Image processing devices such as videophones have been announced.
[0003]
Hereinafter, a conventional image coding method will be described with reference to the drawings. FIG. 5 is a block diagram of a conventional image coding method, FIG. 6 is an explanatory diagram of DCT of the image coding device, FIG. 7 is an explanatory diagram of quantization of the image coding device, and FIG. FIG. 9 is an explanatory diagram of the INTRA / INTER mode of the image encoding device, and FIG. 10 is a stream structure diagram of the image encoding device. Since there are two types of image encoding methods, an INTRA (intra-image encoding) mode and an INTER (inter-image encoding) mode, description will be made according to each case.
[0004]
In FIG. 5, in the INTRA mode, the input image 1 is subjected to DCT by the DCT unit 3, then quantized by the quantization unit 4, and then subjected to variable length coding by the variable length coding unit 5. Then, the encoded information by the encoding control unit 6 and the multiplier unit 7 are multiplied and converted into a transmission bit stream 8.
[0005]
In the case of the INTER mode, the difference between the input image 1 and the predicted image generated by the motion compensation unit 2 is taken, and then the DCT unit 3 performs DCT, and then the quantization unit 4 performs quantization. Next, variable length coding is performed by the variable length coding unit 5, the coding information is coded by the coding control unit 6 and the numbering unit 7, and is converted into a transmission bit stream 8. 11 is a subtracter, 12 and 13 are signal switching means, and 14 is an adder.
[0006]
The data quantized by the quantization unit 4 together with the INTRA mode and the INTER mode is subjected to inverse quantization by the inverse quantization unit 9, inverse DCT by the inverse DCT unit 10, and predicted by the motion compensation unit 2 together with the input image 1. Used to generate an image.
[0007]
FIG. 6 shows the change of the coefficient of the original image 11 by the conventional DCT3. When DCT is applied to the data of the original image 11, it is converted to DCT coefficient 12. When DCT conversion is performed, the data of the original image 11 is scattered almost uniformly, but when converted to the DCT coefficient 12, the absolute value of the component increases toward the upper left and goes to the lower right. The absolute value of the component is getting smaller. At this time, the excessive component at the upper left is called a DC component, the upper left component is called a low frequency component, and the lower right component is called a high frequency component. Furthermore, the low frequency component and the high frequency component are collectively referred to as an AC component. That is, in DCT, the distribution concentrates on the direct current component and the low frequency component, and the value of the high frequency component becomes small.
[0008]
FIG. 7 shows the change of the DCT coefficient 12 by the quantization unit 4. The quantization process is performed based on the quantization characteristic value Q determined by the encoding control unit 6. Quantization is performed separately for a DC component and an AC component. For the direct current component, the component is divided by 2, and the integer part is used as the quantization component. For the AC component, the component is divided by 2 × Q, and the integer part is defined as the quantization component. As a result, all AC components less than 2 × Q are eliminated and become zero. Actually, the DCT coefficient 12 is quantized with Q = 8, and the obtained coefficient is the quantized coefficient 13. Values other than the 6 pixels at the left corner for quantization are 0. Data of the original image 11 is converted into a quantized component of a quantized coefficient 13 having only very few components, and information is compressed.
[0009]
Further, the quantization component of the quantized coefficient 13 whose information amount is compressed and the motion vector generated by the motion compensation unit 2 of 2 are further reduced by the variable length encoding unit 5 and generated by the encoding control unit 6. The control information thus arranged is rearranged in a format determined by the multi-quantization unit 7 to generate a transmission bit stream 8. The transmission bit stream 8 generated here is decoded to the original image by performing the reverse operation at the partner terminal.
[0010]
FIG. 8 is an explanatory diagram of the motion compensation unit 2. This motion compensation is performed when the encoding mode is the INTER mode. First, for a new frame 17 to be encoded, the quantum of the difference image between the immediately preceding frame encoded immediately before this frame is encoded and the frame encoded immediately before it is encoded. The quantization coefficient is subjected to inverse quantization by the inverse quantization unit 9 and inverse DCT by the inverse DCT unit 10 to obtain the sum of the previous predicted image and the same frame as that reproduced on the previous partner side (immediately before The predicted image 14) is reproduced. Thereafter, a motion vector search is performed on the immediately preceding predicted image 14 and the new frame 17 to be encoded this time. The motion vector search is performed according to the following procedure. For the macro block 16 of the new frame 17, the block 15 that matches the immediately preceding predicted image is searched, and the positional deviation from the block at that time is the motion vector 18. The current predicted image is a block in which the block 15 that matches the previous predicted image is pasted at the position of the macro block 16 in the new frame 17.
[0011]
FIG. 9 shows encoded images in the INTRA mode and the INTER mode. In the INTRA frame 19 encoded in the INTRA mode, the generated image itself is used as image data to be encoded. In the INTER mode, the difference between the predicted image and the new frame is taken and used as image data for encoding the image (INTER frame 20). In the INTRA frame 19, the new frame itself is sent, so that the amount of code is very large. In the INTER frame 20, the difference data is generated only in the moving part, so the amount of code is small. With a Q value of 8, the code amount of the INTER frame 20 is reduced to about 1/10 of the code amount of the INTRA frame 19.
[0012]
FIG. 10 shows a sequence of encoding modes at the time of image encoding. In general, in image compression coding, the first frame (frame number 1) that has been coded is coded in the INTRA mode, and the subsequent frames are coded in the INTER mode. Among the frames in the INTER mode, the INTRA mode is inserted in units of macro blocks so that all macro blocks are updated in units of 100 frames.
[0013]
[Problems to be solved by the invention]
In the conventional image compression encoding method, the encoding mode is encoded once in the INTRA mode and then sent out in the INTER mode. This procedure reduces the amount of code, but encodes from the middle of the bitstream when there is an error in the random access when the transmission bitstream is recorded on the medium or when the transmission bitstream is transmitted wirelessly. Although it is necessary to decode the image, if there is only information in the INTER mode, the update to the INTRA mode is performed in units of macroblocks. Therefore, it takes time of 100 frames or more until the entire image is restored. Had. Therefore, in this image compression coding system, it is required to shorten the time for decoding an image from an arbitrary position in the transmission bit stream. If all frames are transmitted as INTRA frames, it is possible to completely restore an image from an arbitrary position in the transmission bitstream. However, a code amount more than 10 times that of an INTER frame is generated. .
[0014]
Therefore, the present invention can shorten the time for decoding the image from an arbitrary position in the transmission bit stream in the moving image compression process, and the amount of generated code is less than that in the case of the INTRA frame. The purpose is to provide a conversion method .
[0015]
[Means for Solving the Problems]
According to the first aspect of the present invention, when performing image coding in the moving image compression coding system, the macroblock units are sequentially coded in the INTRA mode so that the image coding mode becomes a checkered pattern that does not overlap the images. All macroblocks are encoded in INTRA mode twice .
[0018]
DETAILED DESCRIPTION OF THE INVENTION
According to the first aspect of the present invention , an image can be completely restored by reproducing two frames from an arbitrary position in the transmission bit stream.
[0021]
Hereinafter, embodiments of the present invention will be described with reference to the drawings. FIG. 1 is a block diagram of an image compression coding system according to an embodiment of the present invention, and FIGS. 2, 3, and 4 are explanatory diagrams of a switching method for each macroblock in the same coding mode. Since this compression coding system has two types of coding modes, an INTRA (intra-picture coding) mode and an INTER (inter-picture coding) mode, description will be made according to each case.
[0022]
In FIG. 1, reference numeral 21 denotes an encoding control unit that switches the encoding mode according to the method of the present invention. In the INTRA mode, the input image 1 is subjected to DCT by the DCT unit 3, then quantized by the quantization unit 4, and then subjected to variable length coding by the variable length coding unit 5. Furthermore, the encoded information by the encoding control unit 21 that switches the encoding mode according to the method of the present invention and the multiplier 7 are multiplied and converted into a transmission bit stream 8.
[0023]
In the case of the INTER mode, the difference between the input image 1 and the predicted image generated by the motion compensation unit 2 is taken, and then the DCT unit 3 performs DCT, and then the quantization unit 4 performs quantization. Next, variable length coding is performed by the variable length coding unit 5, the number is increased by the encoding control unit 21 and the numbering unit 7, and converted into the transmission bit stream 8. 11 is a subtracter, 12 and 13 are signal switching means, and 14 is an adder.
[0024]
FIG. 2 shows a switching method for each macroblock in the coding mode. The INTRA blocks are sequentially arranged in n frames so that the macroblocks are updated from the top. In FIG. 2, description will be made assuming that n = 3. In a frame where the frame number is divided by 3 and is 0 (denoted as 0 mod n), 1/3 macroblocks from the top of the frame are encoded in the INTRA mode, and the frame number divided by 3 is the 1 frame. In the (1 mod n), 1/3 to 2/3 macroblocks from the top of the frame are encoded in the INTRA mode, and the frame number divided by 3 is 2 frames (denoted as 1 mod n). From the top of the frame, 2/3 to 3/3 macroblocks are INTRA mode decoded. When expressed in general n, when the frame number divided by n is m, frames from m / n to (m + 1) / n are encoded with the INTRA block.
[0025]
By switching the encoding mode to the INTRA mode in this way, the entire screen can be restored only by reproducing n frames. In this case, the code generation amount is about (n + 9) / 10n compared to the case where all frames are continuously sent by INTRA.
[0026]
FIG. 3 shows a switching method for each macroblock in the encoding mode. Place the INTRA / INTER block in a checkered pattern so that the upper left is the INTRA block in the frame where the frame number divided by 2 is 0, and the upper left is the INTER in the frame where the frame number divided by 2 is 1 The INTRA / INTER block is arranged in a checkered pattern so as to become a block. As a result, the entire screen can be restored only by reproducing two frames. In this case, the code generation amount is about (n + 1) / 2n as compared with the case where all frames are continuously sent by INTRA.
[0027]
FIG. 4 shows a switching method for each macroblock in the encoding mode. This is an encoding method in which the central part of the encoded frame is always encoded in the INTRA mode. In this case, the central INTRA block of the image is always restored regardless of the position where the transmission bit stream generated by encoding is decoded. In this case, the code generation amount is about 2/5 as compared with the case where all frames are continuously sent by INTRA.
[0028]
【The invention's effect】
According to the first aspect of the present invention , the image can be completely restored by reproducing two frames from an arbitrary position of the transmission bit stream.
[Brief description of the drawings]
FIG. 1 is a block diagram of an image encoding system according to an embodiment of the present invention. FIG. 2 is an explanatory diagram of a switching method for each macroblock of an encoding mode of an image encoding system according to an embodiment of the present invention. 3 is an explanatory diagram of a switching method for each macroblock of the coding mode of the image coding system according to the embodiment of the present invention. FIG. 4 is a macroblock of the coding mode of the image coding system of the embodiment of the present invention. FIG. 5 is a block diagram of a conventional image coding method. FIG. 6 is a diagram of DCT of a conventional image coding device. FIG. 7 is a diagram of quantization of a conventional image coding device. 8 is an explanatory diagram of motion compensation of a conventional image encoding device. FIG. 9 is an explanatory diagram of an INTRA / INTER mode of the conventional image encoding device. FIG. 10 is a structural diagram of a stream in the conventional image encoding device. Explanation of symbols]
DESCRIPTION OF SYMBOLS 1 Input image 2 Motion compensation part 3 DCT part 4 Quantization part 5 Variable length encoding part 6 Encoding control part 7 Multiplication part 8 Transmission bit stream 9 Inverse quantization part 10 Inverse DCT part 11 Subtractor 19 INTRA frame 20 INTER frame

Claims (1)

When performing image encoding in the moving image compression encoding method, the macroblock units are sequentially encoded in the INTRA mode so that the image encoding mode becomes a checkered pattern that does not overlap the images, and all macros are performed twice. An image encoding method , wherein a block is encoded in an INTRA mode.

Images