Image analysis – Image compression or coding
Reexamination Certificate
2002-01-29
2004-03-02
Tran, Phuoc (Department: 2621)
Image analysis
Image compression or coding
Reexamination Certificate
active
06701018
ABSTRACT:
BACKGROUND OF THE INVENTION
The present invention relates to an image data coding and/or decoding system for compressing and coding an image data into a smaller amount of data and for decoding code strings obtained by the compression coding to reproduce an image. More specifically, the invention relates to an image data coding and/or decoding system which has high error resilience and which can transmit and/or store a coded image data of high quality even if the coded image data is transmitted and/or stored via a medium wherein an error easily occurs, such as a radio channel.
In an image transmitting and/or storing system such as a video telephone, a videoconferencing system, a portable information terminal, a digital video disc system and a television broadcasting system, various systems, which include the motion compensation, the discrete cosine transform, the sub-band coding, the pyramid coding and the combinations thereof, have been developed as techniques for compressing and coding an image data into a smaller amount of data for transmission/storing. In addition, as international standard methods for compressing and coding a moving picture, the methods “ISO•MPEG1”, “ISO•MPEG2”, “ITU-T•H.261”, and “ITU-T•H.262” are standardized. All these methods are compressing and coding methods, each comprising a combination of the motion-compensated adaptive prediction and the discrete cosine transform, and described in detail in “International Standard of Multimedia Coding” (edited and written by Hiroshi Yasuda, published by Maruzen, June 1991) (Literature 1) and so forth.
As an example of conventional moving-picture coding systems, the basic construction of a coding system using the motion-compensated adaptive prediction and the discrete cosine transform is shown in FIG.
1
. In this figure, after an input image signal S
1
is divided into a plurality of regions defined by a region divider
1
, the motion-compensated adaptive prediction is carried out. That is, a motion-compensated adaptive predictor
2
detects a motion vector between the input image signal S
1
and a reference image signal S
2
of the previous frame, which is stored in a frame memory
3
and which has been already coded and locally decoded, and performs the motion compensation with respect to the reference image signal using the detected motion vector, so as to produce a prediction signal. However, in the motion-compensated adaptive predictor
2
, a preferred prediction mode is selected from intraframe codings (prediction signal=0) directly using the motion compensated prediction and the input image signal S
1
, and the prediction signal S
3
corresponding to the prediction mode is outputted.
Then, in a subtracter
4
, the prediction signal S
3
is subtracted from the input image signal S
1
to output a predictive residual signal S
4
. With respect to each of blocks of a predetermined size, the discrete cosine transform (DCT) of the predictive residual signal S
4
is carried out by means of a discrete cosine transformer
5
. The DCT coefficient obtained by the discrete cosine transform is quantized by means of a quantizer
6
. The DCT coefficient quantized by the quantizer
6
is divided into two portions. One of the two portions is coded by means of a variable-length encoder
7
, and then, it is multiplexed with the motion vector, which has been coded by a variable-length encoder
9
, by means of a multiplexer
8
to be outputted as a bit-stream. The other portion is inverse-quantized by means of a inverse quantizer
10
, and then, the inverse discrete-cosine transform (inverse DCT) thereof is carried out by means of an inverse discrete-cosine transformer
11
. The output of the inverse discrete-cosine transformer
11
is added to the adaptive prediction signal S
3
by means of an adder
12
to be a locally decoded signal to be stored in the frame memory
3
.
FIG. 2
is a view illustrating the basic construction of a moving-picture decoding system which corresponds to the moving-picture coding system of FIG.
1
. The code string transmitted from the moving-picture coding system to be stored is divided into a quantized DCT coefficient and a motion vector data by means of a demultiplexer
13
. The quantized DCT coefficient data passes through a variable-length decoder
14
, a inverse quantizer
15
and a inverse discrete-cosine transformer
16
, to be outputted as a predicted error signal S
6
. The motion vector data is decoded by means of a variable-length decoder
17
, and then, it is inputted to a motion-compensation predictor
18
. In the motion-compensation predictor
18
, the motion compensation to a reference image signal S
7
of the last frame in a frame memory
19
is carried out using the motion vector to produce a prediction signal S
8
. Then, in an adder
20
, the predicted error signal S
6
is added to the prediction signal S
8
to reproduce an image signal S
9
. The reproduced image signal S
9
is outputted to the outside of the system and stored in the frame memory
19
as the reference image signal S
7
.
However, in such a conventional moving-picture coding and/or decoding system, there are the following problems.
In a channel in which an error may be mixed, such as a radio channel, when only the aforementioned coding is carried out, the quality of the decoded image is remarkably deteriorated if an error occurs. In particular, when there is an error in a signal such as a synchronizing signal, a mode data and a motion vector, the picture quality is remarkably deteriorated.
In addition, as mentioned above, the motion-compensated adaptive prediction coding is frequently used in the moving picture coding. However, since only the interframe difference is coded in the motion-compensated adaptive prediction coding, when an error occurs, the frame is not only incorrect, but an incorrect image is also stored in a frame memory, so that a predicted image is prepared using the incorrect image and a residual error is added thereto. For that reason, even if the subsequent frame is correctly decoded, it is not possible to obtain a correctly decoded image in the subsequent frame, except that when the data is transmitted in a mode (INTRA mode) wherein the coding is performed in only the frame without using the interframe difference or when the influence of the error is gradually attenuated to return to the original state.
FIG. 3
illustrates this condition. This example shows the condition that a black circle is moving. Signals including a residual signal representative of a circle of the next frame (expressed by the black circle of the residual signal) and a residual signal for erasing a circle of the last frame (expressed by a dotted-line circle of the residual signal) are outputted as interframe differential is signals. In this case, the motion compensation is not performed for simplification, and the interframe difference is obtained assuming that MV (motion vector)=0.
If data of one frame is lost due to error, the second frame is not decoded at all, and for example, the first frame is outputted as it is. In the third frame, a residual error, which is capable of correctly decoding if it is added to the second frame, is added to the first frame to reproduce a quite different image. Thereafter, since the residual error is added to the incorrect image, the error is not basically erased, so that it is not possible to reproduce a correctly decoded image.
In order to eliminate the aforementioned problem, a method called “refresh” for coding in the INTRA mode every a predetermined cycle has been conventionally used. In this case, when the coding is performed in the INTRA mode, the code amount is increased to remarkably deteriorate the picture quality when no error occurs. Therefore, methods such as the cyclic refresh for refreshing every few macroblocks in one frame, not simultaneously refreshing the whole picture, are generally used. However, although the cyclic refresh is able to resstring the increase of the code string, there is a problem in that it takes a long time until the norm
Chujoh Takeshi
Dachiku Kenshi
Kikuchi Yoshihiro
Nagai Takeshi
Watanabe Toshiaki
Finnegan Henderson Farabow Garrett & Dunner L.L.P.
Kabushiki Kaisha Toshiba
Tran Phuoc
LandOfFree
Moving-picture signal coding and/or decoding system... does not yet have a rating. At this time, there are no reviews or comments for this patent.
If you have personal experience with Moving-picture signal coding and/or decoding system..., we encourage you to share that experience with our LandOfFree.com community. Your opinion is very important and Moving-picture signal coding and/or decoding system... will most certainly appreciate the feedback.
Profile ID: LFUS-PAI-O-3233362