JP2869164B2 - Image encoding and decoding device - Google Patents

Description

Description: TECHNICAL FIELD The present invention relates to an image encoding and decoding apparatus that encodes a moving image signal and decodes the encoded image signal.

(Prior Art) As one of communication systems, there is a packet communication system. That is, communication is performed by dividing a transmission code into a certain size, adding a code (header) indicating a destination and contents to each of them, compiling them as packets (cells), and exchanging these cells on a network. That's what I mean. The packet communication system has attracted attention as a next-generation communication system because it can handle voice, image, and data communication processing in a unified and large amount.

However, packet communication systems also have disadvantages. That is, when the total number of cells exceeds the processing capacity of the network due to network congestion, a phenomenon called cell discard occurs. When cell loss occurs,
In the case of image communication, image quality is degraded, for example, a part of an image is lost.

Therefore, in the case where an image is encoded and transmitted by the packet communication method, the image encoding method must be a method that takes this cell discard into account.

Therefore, conventionally, image encoding and decoding in consideration of the packet communication system have been performed by the following methods shown in FIGS. 2 (a) and 2 (b) (for example, literature: Hamano, Sakai, Matsuda:
“A Study on Image Coding Method with Resistance to Cell Discard”,
IEICE, Technical Committee on Image Engineering, Image Coding Symposium PCSJ89 PP97-98 (1989.10), etc.).

First, conventional image coding will be described with reference to FIG.

FIG. 2A is a block diagram showing an example of a conventional image encoder. In the figure, an original image is converted into an electric signal representing a light and shade distribution by scanning of an image pickup tube of a TV camera 201, and further converted into an analog / digital converter (hereinafter, A / D).
It is called a converter. ) 202 to sample into pixels and convert them into digital image signals. The digital image signal is input to a discrete cosine converter (hereinafter, referred to as a DCT converter) 203. The DCT converter 203 converts the input image signal into several processing blocks (for example, a processing block including 16 × 16 pixels).
And obtains a transform coefficient block for each processing block.

The transform coefficient obtained by the DCT converter 203 is input to the adder 204, and the difference between the transform coefficient of the previous frame and the transform coefficient stored in the frame memory 209 is obtained. The residual signal obtained by taking this difference is input to the quantizer (Q) 205. The quantizer 205 quantizes the residual signal (transform coefficient) with, for example, 8 bits, and outputs the quantized transform coefficient to the variable length coder (VLC) 210. The variable length encoder 210 encodes the code of the transform coefficient with an unequal length code such as a Huffman code,
Output to the packet assembling circuit (PAD) 211. The packet assembling circuit 211 divides the output code of the variable-length encoder 210 into an appropriate size, attaches the header, combines the cells into cells, and sends the cells to the packet communication network 212. The quantized transform coefficient output from the quantizer 205 is also input to the inverse quantizer 206, where it is inversely quantized and input to the adder 207.
The adder 207 adds the transform coefficient of one frame before subtracted to the transform coefficient that has been dequantized, and outputs a local decoded signal. The local decoded signal is multiplied by a leak coefficient α by a multiplier 208, and then input to a frame memory (FM) 209 and stored as a prediction signal of the next frame.

Next, conventional image decoding will be described with reference to FIG.

FIG. 2 (b) is a block diagram showing an example of a conventional image decoder. This image decoder corresponds to the image encoder shown in FIG.

In FIG. 7B, the packet disassembly circuit 213 receives cells from the packet communication network 212 and outputs the cells to the variable length decoders 214, respectively. The variable length decoder 214 decodes the unequal length code into a transform coefficient code, and outputs the transform coefficient code to the inverse quantizer 215. The inverse quantizer 215 inversely quantizes the input code into transform coefficients and outputs the result to the adder 216. Adder 216 is
A transform coefficient of the frame memory 218 is added to the inversely quantized transform coefficient, and the decoded signal is output to a discrete cosine inverse transformer (hereinafter, referred to as a DCT inverse transformer) 219 and a multiplier 217. The multiplier 217 multiplies the decoded signal output from the adder 216 by a leak coefficient α, and
Output to The frame memory 218 stores this signal as a prediction signal of the next frame. DCT inverse transformer 219
The reproduced signal (decoded signal) is subjected to orthogonal inverse transform to be converted into a reproduced image signal. The reproduced image signal is thereafter converted into an analog image signal by a digital / analog converter (hereinafter, referred to as a D / A converter) 220 and output to the image monitor 221.

Here, in the encoder of FIG. 2A, a DCT transformer 203 is used as an orthogonal transform, and predictive coding in a transform coefficient domain, that is, an adder 204 calculates an inter-frame difference for a transform coefficient. It is characterized by taking. Further, the encoders and decoders shown in FIGS. 2A and 2B are provided with adders 208 and 217 for multiplying leak coefficients α (0.0 <α <1.0), respectively. In these multipliers 208 and 217, the output of the local decoder of the block 200 portion (the local decoded signal which is the output of the adder 207) indicated by the one-dot chain line in FIG.
The outputs of the adders 216 (the outputs of the adders 207 and 216 are DCT transform coefficients (discrete cosine transform coefficients) to be reproduced) are selectively multiplied by a leak coefficient α to obtain the following (1) and (2). 2), and an encoding / decoding system suitable for packet communication is realized.

(1) By multiplying the DCT transform coefficient by the leak coefficient α, the influence of past frames can be discarded little by little. Therefore, even if the cell is discarded, the reproduced image of the decoder is momentarily disturbed, but is automatically recovered with the passage of time.

(2) The frequency component of the image can be selectively multiplied, such as 0.99 for the conversion coefficient corresponding to the DC component of the image and 0.9 for the conversion coefficient corresponding to the AC component of the image. Therefore, unlike the conventional leak of the inter-frame difference in the time domain, the coding rate does not extremely increase due to the decrease in the prediction gain.

(Problems to be Solved by the Invention) However, the above-described conventional image coding and decoding methods have the following problems.

(1) A DCT converter is applied to an image signal, and a simple inter-frame difference is obtained for the transform coefficient. Therefore, motion compensation prediction coding, which is a prediction coding method in the time domain, cannot be used, resulting in poor coding efficiency.

(2) In the case where leakage is selectively performed to prevent a decrease in prediction gain, a plurality of multipliers must be prepared, and the processing becomes complicated.

Therefore, an object of the present invention is to solve the above-described problems, and to simply perform predictive coding in the time domain, and not to reduce the prediction gain, to achieve high coding efficiency, and even to cause cell discard. Playback image quality automatically recovers over time,
An object of the present invention is to provide an excellent image encoding and decoding apparatus.

(Means for Solving the Problem) In the first invention, a discrete cosine transform is performed on a residual signal between an input image signal and a predicted image signal from a first frame memory, and the transform coefficient is quantized and encoded. In an image coding apparatus having an inter-frame predictive encoder for transmitting an output code, the inter-frame predictive encoder subtracts a predetermined value from a moving image signal and transmits a result as the input image signal. And a local decoder for obtaining a locally decoded image signal from the output code and the predicted image signal from the first frame memory; and a local decoded image signal from the local decoder. A first multiplier that multiplies a leak coefficient and outputs the multiplication result to the first frame memory so as to store the multiplication result in the first frame memory as a predicted image signal of a next frame. It becomes.

According to a second aspect of the present invention, the encoded image signal is inversely quantized, the inverse discrete cosine transform is performed, and a prediction signal from the second frame memory is added to the obtained reproduction residual signal to perform the first reproduction. In an image decoding apparatus having an inter-frame prediction decoder that outputs an image signal, the inter-frame prediction decoder adds a predetermined value to the first reproduced image signal and outputs a second reproduced image signal A second adder,
A second multiplication for multiplying the first reproduced image signal by a leak coefficient and outputting the multiplication result to the second frame memory so as to store the multiplication result as a prediction signal of a next frame in the second frame memory; And a container.

According to a third aspect of the present invention, a residual cosine transform between an input image signal and a predicted image signal from the first frame memory is performed by discrete cosine transform, and the transform coefficient is quantized and coded to output an output code. An image encoding apparatus having a predictive encoder, and an inverse quantization of an encoded image signal supplied from the image encoding apparatus via a communication line, and further performing an inverse discrete cosine transform to obtain a reproduction residual obtained. An image decoding device having an inter-frame predictive decoder for adding a prediction signal from a second frame memory to a signal and outputting a first reproduced image signal; An inter prediction encoding encoder that subtracts a predetermined value from a moving image signal and sends the result as the input image signal; a first adder that outputs the output code and a prediction image from the first frame memory; A local decoder for obtaining a locally decoded image signal, and a local decoded image signal from the local decoder, multiplied by a leak coefficient, and the multiplication result is stored in the first frame memory. A first multiplier that outputs the predicted value to the first frame memory so as to store the predicted value as a predicted image signal of the frame. The inter-frame predictive decoder outputs the predetermined value to the first reproduced image signal. In addition, a second adder for outputting a second reproduced image signal, and a multiplication of the first reproduced image signal by a leak coefficient, and the multiplication result is stored in the second frame memory to predict the next frame. And a second multiplier for outputting the signal to the second frame memory so as to store the signal as a signal.

(Operation) The image encoding device has an inter-frame prediction encoder. The inter-frame predictive encoder subtracts a predetermined value from a moving image signal and sends the result as an input image signal, a local decoder, and a local decoder from the local decoder. A first multiplier that multiplies the converted image signal by a leak coefficient and outputs a result of the multiplication to a first frame memory as a predicted image signal of a next frame. The inter-frame predictive encoder performs a discrete cosine transform of a residual signal between the input image signal from the first adder and the predicted image signal from the first frame memory, and quantizes and encodes the transform coefficient. And sends the output code.

On the other hand, the image decoding device has an inter-frame predictive decoder, and the inter-frame predictive decoder has a second adder and a second multiplier. The inter-frame predictive decoder is supplied from the image encoding device via a communication line. The coded image signal is inversely quantized, and inverse discrete cosine transform is performed, and a prediction signal from the second frame memory is added to the obtained reproduction residual signal to obtain a first reproduction image signal.
Further, in the inter-frame predictive decoder, the second adder adds the predetermined value to the first reproduced image signal to obtain a second reproduced image signal (in the embodiment, corresponds to a final reproduced image signal). ) Is output. In the inter-frame prediction decoder, the second multiplier multiplies the first reproduced image signal by a leak coefficient, and outputs the multiplication result to the second frame memory as a prediction signal of the next frame. I do.

As described above, since the image encoding apparatus performs predictive encoding of an image in the time domain, motion-compensated inter-frame prediction can be performed, and encoding can be performed with high encoding efficiency. Further, the inter-frame predictive encoder of the image encoding apparatus automatically protects a DC component by subtracting a predetermined value (a preset fixed value) from a moving image signal, and selectively detects a leak in an AC component. As a result, the prediction gain does not decrease and the coding efficiency is high. Also, in the case where an image is encoded and transmitted by the packet communication method, the influence of the past frame can be gradually removed by leakage even if the cell is discarded, so that the image quality degradation can be automatically recovered. .

(Example) Next, an example of the present invention will be described with reference to the drawings.

FIG. 1 is a block diagram showing an embodiment of an image encoding / decoding apparatus according to the present invention. FIG. 1 (a) shows an embodiment of an image encoding apparatus according to the present invention, and FIG. () Shows an embodiment of the image decoding apparatus according to the present invention.

First, the image coding apparatus according to the present invention shown in FIG.

The original image is converted into an electric signal representing the light and shade distribution by scanning the imaging tube of the TV camera 101, and further sampled into pixels by the A / D converter 102, and converted into a digital image signal. This input digital image signal is analyzed by an analysis filter (QM
F) Entered in 103. The digital image signals band-divided by the analysis filter 103 are respectively encoded by encoders 104a to 104a
Entered in 104d. Each of the encoders 104a to 104d encodes the digital image signal by an inter-frame predictive encoding method. The codes output from the encoders 104a to 104d are input to the variable length encoders 105a to 105d, respectively.
The variable-length encoders 105a to 105d encode the codes from the encoders 104a to 104d into Huffman encoding and run-length encoding, and output them to the packet assembling circuit 106. Packet assembly circuit 106
Is a circuit for combining output codes of the variable length encoders 105a to 105d into a determined cell length, adding a header, and assembling the cells into cells. The cells assembled by the packet assembling circuit 106 are sent out to the packet communication network 107.

Next, the image decoding apparatus that receives cells from the packet communication network 106 will be described with reference to FIG.

Cells from the packet communication network 107 are separated into variable length codes of each band by a packet decomposing circuit 108. Each of these variable-length codes is a variable-length decoder.
The data is decoded by 109a to 109d and input to decoders 110a to 110d. The decoders 110a to 110d reproduce the band-divided image signals using inter-frame prediction. The reproduced image signals of the respective bands are synthesized into a reproduced image by a synthesis filter (QMF) 111. This reproduced image is converted into an analog image signal by the D / A converter 112 and output to the image monitor 113.

Next, a specific configuration showing one embodiment of the encoders 104a to 104d in FIG. 1A will be described with reference to FIG.

FIG. 3 is a block diagram showing one embodiment of each of the encoders 104a to 104d in FIG. 1 (a). Each encoder 10 shown in FIG.
4a to 104d are constituted by a motion compensation inter-frame predictive encoder as shown in FIG.

In FIG. 3, reference numeral 301 denotes an encoder input terminal to which a digital image signal that has been subjected to band division by the analysis filter 103 is supplied. The digital image signal that has undergone band division is input from the encoder input terminal 301. Then, the adder 302
As a result, a certain fixed value A is subtracted from the input image signal. Next, the motion vector detector (MV) 311 performs block matching between the input image signal and the image signal of the frame memory 310 using, for example, a processing block of 16 × 16 pixels to detect a motion vector. The motion vector detected by the motion vector detector 311 is output to the frame memory 310 and the encoder output terminal 312.
The adder 303 calculates the difference between the image signal (predicted value) based on the motion vector from the frame memory 310 and the input image signal, and outputs the obtained residual signal to the DCT converter 304. DCT
Transformer 304 performs a discrete cosine transform on the residual signal using, for example, a processing block of 8 × 8 pixels, and outputs the transform coefficient to quantizer 305. The quantizer 305 quantizes and encodes the transform coefficient, and outputs an output code to the encoder output terminal 312 and the inverse quantizer 306. The inverse quantizer 306 is
The code is reproduced into an inverse quantization value, and output to the DCT inverse transformer 307.

The DCT inverse transformer 307 inversely transforms the reproduced transform coefficient into a reproduced residual signal. The reproduction residual signal obtained by the DCT inverse transformer 307 is added with the image signal (prediction value) from the frame memory 310 by an adder 308 to become a locally decoded image signal. The multiplier 309 multiplies the locally decoded image signal output from the adder 308 by a leak coefficient α (0 <α <1.0). This leakage allows the influence of past frames to be discarded little by little, and in the event of cell loss, automatic recovery of image quality degradation (recovery of synchronization between encoder and decoder) becomes possible.

The output signal of the multiplier 309 is stored in the frame memory 310 as a prediction signal of the next frame.

The inverse quantizer 306, the DCT inverse transformer 307, and the image signal (predicted value) from the frame memory 310 and the DCT inverse transformer 307
The adder 308 that adds the residual signal from the adder 308 constitutes a local decoder 313.

Next, a specific configuration of each of the decoders 110a to 110d in FIG. 1B will be described with reference to FIG.

FIG. 4 is a block diagram showing one embodiment of each of the decoders 110a to 110d in FIG. 1 (b). Each decoder 110a to 110d
Is constituted by a motion compensation inter-frame prediction decoder as shown in FIG.

Reference numeral 401 denotes a corresponding variable length decoder 109i shown in FIG.
A decoder input terminal to which an output code (i is a corresponding one of a to d) is supplied, wherein the decoder input terminal
When the output code is supplied to 401, the inverse quantizer 402
The code is reproduced to an inverse quantization value, and output to the DCT inverse transformer 403. Further, the motion vector from the decoder input terminal 401 is sent to the frame memory 406.

The DCT inverse transformer 403 inversely transforms the inversely quantized transform coefficient into a reproduction residual signal. The reproduction residual signal obtained by the DCT inverse transformer 403 is added with a prediction value based on the motion vector from the frame memory 406 by the adder 404, and becomes a reproduced image signal. The reproduced image signal output from the adder 404 is added to the adder 407.
And supplied to the multiplier 405. The reproduced image signal is added with a certain fixed value A in the adder 407 as in the case of the adder 302 of the encoder shown in FIG. Are output to the synthesis filter 111 of FIG. 1 (b).

Further, the multiplier 405 multiplies the reproduced image signal from the adder 404 by the leak coefficient α, which is the same as the case of the multiplier 309 of the encoder in FIG. About this leakage,
The same can be said for leakage in the encoder of FIG. The output of the multiplier 405 is input to the frame memory 406 as a prediction signal of the next frame.

3 and 4, the adder 302,
The operation of subtracting or adding the fixed value A performed in 407 means removing or adding an important DC component in the image. For this reason, the leak of the encoder of FIG. 3 and the decoder of FIG. 4 is mainly performed on the AC component of the image signal. Therefore, the prediction gain does not significantly deteriorate. Here, the fixed value A is preferably set in the vicinity of the average value of the input image signal in consideration of this. For example, if the input image signal fluctuates uniformly in the range of 8 bits, 0 to 255, if each fixed value A of the encoder of FIG. 3 and the decoder of FIG. 4 is set to 128, Good. Thus, if the same fixed value A is set in each of the encoder and the decoder, it is not necessary to transmit a special parameter from the encoder to the decoder.

The present invention is not limited to the present embodiment, and various applications and modifications can be considered without departing from the gist of the present invention. For example, in the present embodiment, the input image signal is divided into a plurality of bands by using the analysis filter 103, and encoding and decoding are performed in each band. However, the present invention is not limited to this. Without dividing the input image signal into a plurality of bands by the analysis filter 103,
103 and the synthesis filter 111, the encoder and the decoder are replaced by a plurality of encoders 104a to 104d and a plurality of decoders 110a to 110d as shown in FIGS. 1 (a) and 1 (b), respectively. It is also possible to provide one image signal as it is and not to provide the image signal as it is.

(Effects of the Invention) As described above, the present invention has the following various effects.

(1) Since predictive coding of an image is performed in the time domain, motion-compensated inter-frame prediction can be performed, and coding can be performed with high coding efficiency.

(2) The image coding apparatus has an inter-frame predictive encoder. In the inter-frame predictive encoder, a DC component is subtracted from a moving image signal by subtracting a predetermined value (a fixed value set in advance). Since the protection is performed automatically and the AC component is selectively leaked, the prediction gain does not decrease and the coding efficiency is high.

(3) In the inter-frame prediction encoder of the image encoding device, the value subtracted from the input image signal is a predetermined value (a fixed value set in advance), and in the inter-frame prediction decoder of the image decoding device, Since the value added to the reproduced image signal is a predetermined value (a fixed value set in advance), there is no need to exchange special parameters between the image encoding device and the image decoding device, and the process is simplified.

(4) In the inter-frame predictive encoder of the image encoding device or the inter-frame predictive decoder of the image decoding device, the influence of the past frame can be discarded little by little similarly to the related art due to leakage. For this reason, even when the cell is discarded, the image quality deterioration can be automatically recovered (it is strong against the cell discard). Also, the leakage processing in the inter-frame predictive encoder and the inter-frame predictive decoder requires only one multiplier, and the processing is simple.

BRIEF DESCRIPTION OF THE DRAWINGS FIG. 1 is a block diagram showing an embodiment of the present invention, FIG. 2 is a block diagram showing a conventional example, and FIG. 3 and FIG. It is a block diagram which shows each Example of a decoder. 104a to 104d: encoder, 110a to 110d: decoder, 30
2,303,308,404,407 ... Adder, 304 ... DCT converter, 305
…… Quantizer, 306,402 …… Dequantizer, 307,403 …… DC
T-inverter, 309, 405 Multiplier, 310, 406 Frame memory, 311 Motion vector detector, 313 Local decoder.

────────────────────────────────────────────────── ─── Continued on the front page (56) References JP-A-62-125787 (JP, A) JP-A-63-73786 (JP, A) JP-A-2-254886 (JP, A) (58) Field (Int.Cl. ⁶ , DB name) H04N ^7/ 24-7/68

Claims (3)

(57) [Claims] An inter-frame predictive coding for performing a discrete cosine transform of a residual signal between an input image signal and a predicted image signal from a first frame memory, quantizing and encoding the transform coefficient, and transmitting an output code. In the image coding apparatus having the encoder, the inter-frame predictive encoder subtracts a preset fixed value from a moving image signal, and sends a result as the input image signal, a first adder, A local decoder for obtaining a locally decoded image signal based on the code and the predicted image signal from the first frame memory; and a local decoding image signal from the local decoder multiplied by a leak coefficient. A first multiplier that outputs the multiplication result to the first frame memory so as to store the result of the multiplication in the first frame memory as a predicted image signal of the next frame. Image coding device. 2. An image signal obtained by inversely quantizing an encoded image signal is subjected to inverse discrete cosine transform, and a prediction signal from a second frame memory is added to an obtained reproduction residual signal to convert a first reproduction image signal. An image decoding apparatus having an inter-frame predictive decoder for outputting, wherein the inter-frame predictive decoder adds a preset fixed value to the first reproduced image signal and outputs a second reproduced image signal A second adder, which multiplies the first reproduced image signal by a leak coefficient, and stores the multiplied result in the second frame memory as a prediction signal of a next frame. An image decoding device, comprising: a second multiplier that outputs to a memory. 3. An inter-frame predictive encoding for performing a discrete cosine transform on a residual signal between an input image signal and a predicted image signal from a first frame memory, quantizing and encoding the transform coefficient, and transmitting an output code. A coding apparatus having an encoder, an inverse quantization of an encoded image signal supplied from the image encoding apparatus via a communication line, and an inverse discrete cosine transform. In the inter-frame predictive coding / decoding device which outputs a first reproduced image signal by adding a predictive signal from the frame memory, the inter-frame predictive encoder subtracts a preset fixed value from a moving image signal. A first adder for transmitting the result as the input image signal; and a local decoder for obtaining a locally decoded image signal from the output code and the predicted image signal from the first frame memory. And a decoder for multiplying the locally decoded image signal from the local decoder by a leak coefficient, and storing the multiplication result in the first frame memory as a predicted image signal of a next frame. A first multiplier for outputting to the first frame memory, the inter-frame predictive decoder adds the preset fixed value to the first reproduced image signal, and outputs a second reproduced image signal. And a second adder that outputs the first reproduced image signal by a leak coefficient, and stores the multiplied result in the second frame memory as a prediction signal of the next frame. And a second multiplier for outputting the image data to the frame memory.