JPH08116539A - Dynamic image coder and dynamic image coding method - Google Patents

Abstract

PURPOSE: To reduce more the capacity of a frame memory of the dynamic image coder. CONSTITUTION: A differential device 10 calculates a difference between an input signal S0 and an expanded image signal S9 of one preceding frame outputted from an expansion device 90 to provide an output of a signal S1. A frequency converter 20 decomposes an inter-frame difference into a frequency component, and a quantizer 30 divides a range in which transformation coefficients are taken to a finite number and provides a label representing to which the transformation coefficient belongs. An inverse quantization device 50 restores the transformation coefficient into a waveform signal S5. An adder 60 adds the waveform signal and the expanded image signal S9 to output a local decoding signal S6. A compressor 70 compresses the signal S6 to provide an output to a frame memory 80 and the memory 80 delays the stored compressed signal S7 for one frame time and outputs the signal S8. An expander 90 expands the compressed image signal S8 to output an expanded image signal S9 of one preceding frame time.

Description

Detailed Description of the Invention

[0001]

BACKGROUND OF THE INVENTION 1. Field of the Invention The present invention relates to a moving picture coding method and apparatus for compressing a moving picture signal such as a television signal, and more particularly, to a moving picture coding method suitable for reducing the frame memory capacity. Regarding the device.

[0002]

2. Description of the Related Art Since an image signal has an enormous amount of information as compared with an audio signal, compression of the amount of information is applied to its transmission. A moving image used in television broadcasting or the like is composed of about 30 frames called frames per second. When the image signal to be transmitted is a moving image, the inter-frame coding for transmitting the change between the signal value of the frame to be currently transmitted and the signal value one frame before is excellent in terms of compression rate.

To give a numerical value regarding the storage capacity of the frame memory, ITU-T (formerly CCITT) standard H.261 (px 64 kb / s
In the video coding method for audiovisual services), it is necessary to store one frame of image signal, which is approximately 360 (horizontal pixels) × 280 (vertical pixels) × 1.5.
(Luminance signal Y: Color difference signal C1: Color difference signal C2 = 1.0: 0.25: 0.
25) = 150 kBytes, ISO standard MPEG (moving image coding method for storage) requires the storage of 2 frames of image signals, and about 360 × 280 × 1.5 × 2 = 300 kBytes, which is currently urgently recommended. High-definition television signal (HDTV) with MPEG2
When compressing, about 1920 × 1035 × 1.5 × 2 = 6 MB
It becomes the size of ytes.

In the system of the moving picture coding apparatus, in addition to a frame memory for calculating a difference value for interframe coding, a frame memory which is used in various ways is mounted. The apparatus described in Japanese Patent Application Laid-Open No. 2-296484 has a frame memory for accumulating a noise component at the time of compression for one frame in order to improve the image quality of the intraframe coding apparatus. The device described in Japanese Patent Laid-Open No. 3-211985 has a frame memory that stores one frame as a videophone hold signal.

[0005]

In order to popularize a system for transmitting moving images such as a videophone, it is necessary to reduce the manufacturing cost and downsize the device. for that reason,
It is necessary to reduce the capacity of the frame memory, which increases the cost and space of the moving picture coding apparatus. However, the above-described conventional moving picture coding apparatus does not reduce the storage capacity of the frame memory for calculating the difference value for interframe coding.

It is an object of the present invention to provide a moving picture coding method and apparatus capable of reducing the frame memory capacity.

[0007]

SUMMARY OF THE INVENTION The above-mentioned object is to compress an image signal in a preceding stage of a frame memory for calculating a difference value for interframe coding by a compression means, and to read an image signal from the frame memory. By stretching the
Achieved.

[0008]

The compression means compresses the image signal before inputting it to the frame memory to reduce the information amount of the image signal and then inputs it to the frame memory. Therefore, the capacity of the frame memory can be reduced as compared with the conventional moving picture coding apparatus using interframe coding.

[0009]

DETAILED DESCRIPTION OF THE PREFERRED EMBODIMENTS An embodiment of the present invention will be described below with reference to the drawings. FIG. 1 is a block diagram of a moving picture coding apparatus according to an embodiment of the present invention. 10 is a differentiator, 20 is a frequency converter, 30 is a quantizer, 40 is an inverse quantizer, 50 is a frequency inverse transformer, 60 is an adder, 70 is a compressor, 80 is a frame memory, and 90 is a decompressor. Is. Further, S0 is an input signal, S1 is an output signal of the differentiator 10, S3 is an output signal of the quantizer 30, S5 is an output signal of the frequency inverse converter 50, S6 is a locally decoded signal, and S7 is a compressed locally decoded signal. , S8 is a compressed image signal output from the frame memory 80, and S9 is an expanded image signal of one frame time before.

The differentiator 10 calculates a difference value, that is, a change amount between the input signal S0, that is, the current frame and the decompressed image signal S9 output by the decompressor 90 one frame before, and outputs it as a signal S1. The frequency converter 20 decomposes the inter-frame difference value into frequency components, that is, conversion coefficients. As a frequency conversion method, for example, discrete cosine transform (DCT: DiscreteCo
sine Transform) is adopted. The quantizer 30 omits, from the plurality of frequency-transformed transform coefficients that are less noticeable to human vision, that is, sets the transform coefficient to zero, divides the range of possible values of the transform coefficient into a finite number, and divides the range. Label the signal to indicate which range the transform coefficient belongs to
It is output to an entropy encoder (not shown) as S3. The inter-frame difference value output to the entropy encoder is a redundant signal that generally has a large probability distribution in a small portion near zero. Using this property, the entropy coder specifically compresses the code length of the image signal by Huffman coding or run-length coding.

The inverse quantizer 40 returns the label given by the quantizer 30 to the signal value of the transform coefficient, and the frequency inverse transformer 50 returns the transform coefficient to the waveform signal and outputs it as the signal S5. The adder 60 receives the waveform signal S5 and the expander 90.
The image signal S9 decompressed in step S6 is added and the local decoded signal S6 is output. The compressor 70 compresses the locally decoded signal S6,
The compressed locally decoded signal S7 is output to the frame memory 80. The frame memory 80 is a compressed local decoded signal.
After delaying S7 for one frame time, it is output as a signal S8. The decompressor 90 decompresses the compressed image signal S8 and outputs it as the decompressed image signal S9 one frame time ago. In this way, by providing the compressor 70 in the previous stage of the frame memory 80, interframe coding can be realized with a frame memory having a smaller capacity than in the case where no compression is performed.

Note that the present invention is not limited to interframe coding using frequency conversion, and basically any interframe coding method for comparing the current pixel signal with the signal one frame time before can be used. It can be applied to both cases.

FIG. 2 is a detailed block diagram of the compressor 70 shown in FIG. Reference numeral 71 is a differentiator, 75 is a one-pixel time delay element such as a D-type flip-flop, and 76 is a Huffman encoder. Further, S6 is a locally decoded signal, S7
Is the compressed locally decoded signal.

The compressor 70 shown in FIG. 2 does not have a quantizer and executes lossless encoding in which the compressed code is always restored to the original image signal. The differentiator 71 takes the difference value between the locally decoded signal S6 and the locally decoded signal delayed by the delay element 75 for one pixel time, and the Huffman encoder 76 compresses the code length.

FIG. 3 is a detailed block diagram of the expander 90 shown in FIG. 91 is a Huffman decoder, 93 is an adder, 9
Reference numeral 4 is a delay element for one pixel time, such as a D type flip-flop. Further, S8 is an as-compressed image signal delayed by one frame time, and S9 is an expanded image signal.

The decompressor 90 shown in FIG. 3 corresponds to the compressor 70 which executes lossless coding in which the compressed code is always restored to the original image signal. The Huffman decoder 91 decodes the as-compressed image signal S8 delayed by one frame time and restores it to the difference value between adjacent pixels. The adder 93 adds the image signal one pixel before, which is output by the delay element 94 for one pixel time, and the current difference value, which is output by the Huffman decoder 90. The current image signal is an expanded image signal. Output as S9. With lossless encoding,
The expanded image signal S9 is obtained by delaying the locally decoded signal S6 by one frame time, and new deterioration of the image signal due to the application of this embodiment does not occur. Therefore, IT
UT (formerly CCITT) standard H.261 (px 64kb / s video coding method for audiovisual services), ISO standard MPE
Compatibility with standardized methods such as G (video encoding method for storage) can be ensured.

FIG. 4 is a block diagram of a compressor 70 according to another embodiment. 71 is a differentiator, 72 is a quantizer, 73 is an inverse quantizer, 74 is an adder, 75 is a delay element of one pixel time such as a D-type flip-flop, and 76 is a Huffman encoder. Further, S6 is a locally decoded signal, S7
Is the compressed locally decoded signal.

The compressor 70 shown in FIG. 4 includes a quantizer,
The compressed code is an irreversible coding that does not necessarily return to the original image signal. The differentiator 71 takes the difference value between the locally decoded signal S6 and the image signal output by the delay element 75 for one pixel time, quantizes it by the quantizer 72, and then compresses the code length by the Huffman encoder 76. To do. The signal output from the quantizer 72 is also input to the inverse quantizer 73, and the label given by the quantizer 72 is returned to the signal value.
The adder 74 adds the image signal output by the 1-pixel time delay element 75 and the signal output by the inverse quantizer 73, and inputs the result to the 1-pixel time delay element 75.

When the lossy coding is applied, compatibility with the standardized system cannot be secured, but there is an effect that the compression rate is improved and the frame memory can be further reduced. Further, when the encoding device of the present embodiment and the standardization type encoding device are connected, a screen mismatch occurs between the transmitting side encoding device and the receiving side decoding device. In this case, refresh control may be performed so that the contents of the frame memories of the transmitting side encoding device and the receiving side decoding device are made to coincide with each other on a regular basis.

FIG. 5 is a block diagram of an expander 90 according to another embodiment. Reference numeral 91 is a Huffman decoder, 92 is an inverse quantizer, 93 is an adder, and 94 is a delay element for one pixel time, such as a D-type flip-flop. Also, S
Reference numeral 8 is an as-compressed image signal delayed by one frame time, and S9 is an expanded image signal.

The decompressor 90 shown in FIG. 5 corresponds to the compressor 70 which performs lossy coding in which the compressed code is not always restored to the original image signal. The Huffman decoder 91 decodes the as-compressed image signal S8 delayed by one frame time and restores it to the difference value between adjacent pixels. The inverse quantizer 73 converts the difference value between adjacent pixels into a level that matches the signal unit. Adder 93
Is an addition of the image signal one pixel time before output by the delay element 94 of one pixel time and the current difference value output by the Huffman decoder 90, and outputs it as the current image signal, that is, the expanded image signal S9. .

In this embodiment, an example of the compressor connected to the front stage of the frame memory is shown in FIG. 2 or 4, and an example of the expander connected to the rear stage of the frame memory is shown in FIG. 3 or FIG. However, the present invention is not limited to the compressors and decompressors shown in these drawings, and can be applied as long as it has similar compression and decompression functions.

FIG. 6 is a diagram for explaining the timing of compression encoding. Input signal S0, output signal S1 of the subtractor 10,
The output signal S3 of the quantizer 30, the output signal S5 of the frequency inverse converter 50, the local decoded signal S6, and the compressed local decoded signal S7 process image signals having the same frame number.
Then, the compressed image signal S8 output by the frame memory 80 and the expanded image signal S9 of one frame time before which is output by the expander 90 process the image signals of the same frame number. The image signals flowing through the signal lines S8 and S9 are compared with the image signals flowing through the signal lines S1, S3, S5, S6 and S7.
It is delayed by one frame time.

The frame memory information amount of FIG.
0 shows the accumulated information amount of the compressed locally decoded signal S7 output by 0 for one frame period. Since the cumulative amount of information does not reach the "Full" which is the maximum capacity of the frame memory, normal compression encoding can be continued.

FIG. 7 is a diagram for explaining the timing of the memory overflow process. Each signal in the figure is the same as that shown in FIG. The frame memory information amount of FIG. 7 indicates the accumulated information amount of the compressed locally decoded signal S7 output by the compressor 70 for one frame period. In this FIG. 7, unlike the case shown in FIG. 6, the accumulated information amount reaches the maximum capacity “Full” of the frame memory, causing an overflow. Therefore, when an overflow is detected, the frame memory is initialized. For example, the output of the decompressor 90 is forcibly set to zero, and one new frame of the input signal S0 is input to the frame memory 80 as the storage content. This new one frame is also transmitted to the receiving side decoding device via a Huffman encoder (not shown in FIG. 1) and a transmission line (not shown) in FIG. 1, and is also stored in the frame memory of the receiving side decoding device as stored contents. input.

FIG. 8 is a diagram for explaining the initialization of the frame memory. The configuration means is the same as that of FIG.
The difference is that 0 is added.

"Full" which is the maximum capacity of the frame memory
When the cumulative information amount reaches, the information amount observing means 150 detects the overflow and outputs it as the control signal S15 to the decompressor 90. The decompressor 90 outputs zero during the next one frame period when the overflow is detected, and otherwise outputs a decompressed image signal.

FIG. 9 is a block diagram of a moving picture coding apparatus according to the second embodiment of the present invention. 10 is a differentiator, 20 is a frequency converter, 30 is a quantizer, 40 is an inverse quantizer, 50 is a frequency inverse transformer, 60 is an adder, 70 is a compressor, 80 is a frame memory, and 90 is a decompressor. , 100 is a second frame memory, 110 is a motion detector, and 120 is a variable delay device. Further, S0 is an input signal, S1 is an output signal of the differentiator 10, S3 is an output signal of the quantizer 30, S5 is an output signal of the frequency inverse converter 50, S6 is a locally decoded signal, and S7 is a compressed locally decoded signal. , S8 is an as-compressed image signal delayed by one frame time, S9 is an expanded image signal, S1
0a is an image signal used for motion detection, S10b is an image signal to the variable delay device 120, S11 is a signal indicating the amount of motion, and S1 is
Reference numeral 2 is an image signal delayed by about 1 frame in consideration of variable delay.

The subtractor 10 calculates a difference value, that is, a change amount between the input signal S0, that is, the current frame and the image signal S12 output from the variable delay device 120 one frame before, and calculates the signal line S.
Output to 1. The frequency converter 20 decomposes the inter-frame difference value into frequency components, that is, conversion coefficients. As the frequency conversion method, discrete cosine conversion is adopted in this embodiment, but other methods may be adopted. The quantizer 30 omits the conversion coefficient from the ones that are inconspicuous to human vision among the plurality of frequency-converted conversion coefficients, that is, sets the conversion coefficient to zero, divides the range of the value of the conversion coefficient into a finite number, and divides the range. Label the range to which the conversion coefficient belongs. Then, the signal is output to an entropy encoder (not shown) via the signal line S3. The inter-frame difference value output to the entropy encoder is a redundant signal that generally has a large probability distribution in a small portion near zero. Using this property, the entropy coder specifically compresses the code length of the image signal by Huffman coding or run-length coding.

The inverse quantizer 40 returns the label given by the quantizer 30 to the signal value of the transform coefficient, and the frequency inverse transformer 50 returns the transform coefficient to the waveform signal and outputs it to the signal line S5. The adder 60 adds this waveform signal and the image signal S12 of about one frame before, which is output from the variable delay unit 120, and outputs a locally decoded signal S6.

The compressor 70 compresses the locally decoded signal S6,
The compressed locally decoded signal S7 is output. The frame memory 80 delays the compressed locally decoded signal S7 by one frame and outputs it to the signal line S8. The decompressor 90 decompresses the compressed image signal S8 and outputs it as the decompressed image signal S9 of one frame time ago. Up to this point, the operation is the same as that of the embodiment shown in FIG.

The present embodiment is characterized in that motion detection and motion compensation are performed. In the moving image compression coding for performing motion detection and motion compensation, a search is performed from where in the previous frame the current pixel signal has moved. Therefore, the search target range must be expanded in the image signal, not in the compressed form. Therefore, the second frame memory 100 stores the expanded image signal in the search target range, and outputs the image signal S10a used for motion detection as appropriate. The motion detector 110 compares the input signal S0 with the image signal S10a and outputs a signal S11 indicating the amount of motion. The variable delay unit 120 adds a necessary variable delay amount to the expanded image signal S10b from the second frame memory 100 in accordance with the signal S11 indicating the motion amount, and the signal line S12
Output to.

According to the present embodiment, it is possible to further reduce the amount of frame memory required in the coding device that performs motion detection and motion compensation.

FIG. 10 is a block diagram of a moving picture coding apparatus according to the third embodiment of the present invention. 200 is an integrated circuit, 20
1 is a central processing unit (CPU), 202 is a frequency converter, 2
Reference numeral 03 is an internal bus, 204 is a compression image bus, 205 is a decompression image bus, 206 is a compressor, 207 is a frame memory, and 208 is a decompressor.

In the present embodiment, the difference unit shown in FIG.
The calculation processing of the quantizer, dequantizer, and adder is performed by the CPU.
201 does. The CPU 201 can also perform the arithmetic processing of the frequency converter and the frequency inverse converter, but since the load of the frequency conversion processing is generally heavy, the frequency converter 206 executes it.

A signal corresponding to the local decoded signal S6 output from the adder 60 shown in FIG. 1 is input to the compressor 206 via the internal bus 203 and the compression image bus 204. The compressor 206 compresses the same signal and stores it in the frame memory 207. The frame memory 207 delays the compressed image signal by one frame and outputs the delayed image signal. The decompressor 208 decompresses the image signal, and the decompressing image bus 205 and the internal bus 203.
Via the CPU 201 or the frequency converter 206. Also in the embodiment shown in FIG. 10, the same effect as in the first and second embodiments can be obtained.

[0037]

According to the present invention, it is possible to obtain a small-sized and low-cost moving image coding apparatus with a smaller frame memory capacity.

[Brief description of drawings]

FIG. 1 is a configuration diagram of a moving picture coding apparatus according to a first embodiment of the present invention.

FIG. 2 is a detailed configuration diagram of the compressor shown in FIG.

FIG. 3 is a detailed configuration diagram of the expander shown in FIG.

FIG. 4 is a configuration diagram according to another embodiment of the compressor.

FIG. 5 is a configuration diagram according to another embodiment of the expander.

[Fig. 6] Fig. 6 is a diagram for describing the timing of compression encoding.

FIG. 7 is a diagram illustrating the timing of memory overflow processing.

FIG. 8 is a diagram illustrating initialization of a frame memory.

FIG. 9 is a configuration diagram of a moving picture coding apparatus according to a second embodiment of the present invention.

FIG. 10 is a configuration diagram of a moving picture coding apparatus according to a third embodiment of the present invention.

[Explanation of symbols]

10 ... Difference device, 20 ... Frequency converter, 30 ... Quantizer,
40 ... Inverse quantizer, 50 ... Frequency inverse transformer, 60 ... Adder, 70 ... Compressor, 80 ... Frame memory, 90 ... Decompressor.

 ─────────────────────────────────────────────────── ─── Continued Front Page (72) Inventor Kozo Nakamura 7-1, 1-1 Omika-cho, Hitachi-shi, Ibaraki Hitachi Ltd. Hitachi Research Laboratory

Claims (11)

[Claims] 1. A moving picture coding apparatus for performing inter-frame coding for comparing a current picture signal with a picture signal one frame time before, the picture coding apparatus being provided in a preceding stage of a frame memory for generating a delay of one frame time. And a video signal decompressing unit provided at a subsequent stage of the frame memory. 2. The moving picture coding apparatus according to claim 1, wherein the image signal compressing means and the image signal expanding means perform lossless coding. 3. The moving picture coding apparatus according to claim 1, wherein the image signal compression means and the image signal decompression means perform lossy coding. 4. The moving picture coding apparatus according to claim 3, further comprising means for periodically refreshing the frame memory. 5. The moving picture coding apparatus according to claim 1, further comprising means for observing the amount of image information stored in the frame memory. 6. The means according to claim 5, further comprising means for making the output signal of the image signal expansion means zero when observing that the amount of image information stored in the frame memory exceeds the storage capacity of the frame memory. And a moving picture coding device. 7. The moving picture coding apparatus according to claim 1, further comprising a second frame memory for previously expanding and storing a part of the entire screen. 8. A central processing means, a frame memory for generating a delay of one frame time, an image signal compression means connected before the frame memory, and an image signal expansion means connected after the frame memory. A moving image having a first signal bus connected to the central processing means and the image signal compression means, and a second signal bus connected to the central processing means and the image signal expansion means. Image coding device. 9. An image to be stored in a frame memory for generating a delay of 1 frame time in a moving image coding method for performing inter-frame coding for comparing a current image signal and an image signal of 1 frame time before. A moving picture coding method, characterized in that a signal is compressed and an output signal is expanded from the frame memory. 10. The moving image coding method according to claim 9, wherein compression and expansion of the image signal are performed by either lossless coding or lossy coding. 11. The moving picture coding method according to claim 10, wherein the frame memory is periodically refreshed when the compression and the expansion are performed by the lossy coding.