Image analysis – Image compression or coding – Interframe coding
Reexamination Certificate
2000-03-09
2004-07-13
Boudreau, Leo (Department: 2721)
Image analysis
Image compression or coding
Interframe coding
C382S232000, C382S251000
Reexamination Certificate
active
06763138
ABSTRACT:
BACKGROUND OF THE INVENTION
1. Technical Field of the Invention
The present invention relates to a method and apparatus for coding moving pictures at a variable bit rate, and more particularly to a method and apparatus for coding moving pictures at a variable bit rate, which can control a code quantity at a variable bit rate in real time.
2. Description of the Prior Art
MPEG-2 (ISO-IEC-13818-2) has been typically known as a method of coding moving pictures with a high efficiency in the art. In this method, a picture is divided into blocks each consisting of a group of plural pixels, and each block is subject to the discrete cosine transform (DCT) to transform a signal in a spatial domain into a signal in a frequency domain. Each frequency component obtained from the discrete cosine transform, which is called a “DCT coefficient”, is quantized with a predetermined quantization scale, then coded in a variable length by assigning a variable length code to a quantized transform coefficient, and output as a bit stream of coded data.
FIG. 10
is a block diagram showing an arrangement of a conventional moving picture coding method for explaining an moving picture coding method on the basis of MPEG-2 (ISO-IEC-13818-2). The moving picture coding method comprises, as shown in
FIG. 10
, subtracter
501
, discrete cosine transform (DCT) means
502
, quantization means
503
, variable length coder
504
, bit rate controller
505
, inverse quantization means
506
, inverse DCT means
507
, adder
508
, frame memory
509
, motion compensation interframe predictor
510
and selector
511
.
Input images are classified into an I-picture which is encoded with intra-frame coding, a P-picture which is encoded with interframe prediction coding with only a forward prediction, and a B-picture which is encoded with bi-directional or forward and backward predictions. The input images are divided into macro blocks each consisting of 16×16 pixels and are coded on macro block basis.
First, in the I-picture, the intra-frame coding is performed without interframe prediction. At this moment, selector
511
selects a signal value “0” as a predicted value, and subtracter
501
outputs the same value as the signal value of the input image. An output signal value from subtracter
501
is transformed through the discrete cosine transform at DCT means
502
into a DCT coefficient, which is outputted toward quantization means
503
. The DCT coefficient is quantized at quantization means
503
with a predetermined quantization scale.
The DCT coefficient quantized at quantization means
503
is output as a quantized transform coefficient to variable length coder
504
, where the quantized transform coefficient is coded in a variable length. The quantized transform coefficient coded in the variable length is output as a bit stream along with other data. The quantized transform coefficient output from quantization means
503
is also fed to inverse quantization means
506
to reconstruct original image signals.
The image signals reconstructed through the inverse quantization means
506
and the inverse DCT means
507
will be employed as a reference picture on the subsequent interframe prediction, and are stored in the frame memory
509
through the adder
508
. The interframe prediction is performed to P-picture and B-picture at the motion compensation interframe predictor
510
.
The motion compensation interframe predictor
510
first compares the input image signals with the reference pictures stored in the frame memory
509
, and performs a motion vector detection to determine a motion vector on every macro block that is obtained from dividing the input image. A prediction mode is determined per macro block on the basis of the result from the motion vector detection. In this process, it is decided whether a process of coding a macro block for P-picture and B-picture is executed preferably in the intra-mode or not.
If the prediction mode for a macro is the intra-mode, then the process of coding the macro block for P-picture and B-picture is also performed in an intra-frame coding process. On the contrary, if the prediction mode for a macro block is not the intra-mode, then the process of coding the macro block for P-picture and B-picture is perfomed by the interframe prediction using the image signals stored in the frame memory
509
as reference images. The motion compensation interframe predictor
510
performs a motion compensation interframe prediction based on the preveously determined motion vector to generate a predicted image signal corresponding to the input image signal.
The predicted image signal generated from the motion compensation interframe predictor
510
is given through the selector
511
to the adder
508
as well as the subtracter
501
. The predicted image signal given to the subtracter
501
is herein described. The subtracter
501
computes a difference signal between the input image signal and the given predicted image signal. A coding process subsequently performed to the difference signal is the same as the coding process to I-picture; the difference signal is transformed at the DCT means
502
into a DCT coefficient, then is quantized at the quantization means
503
.
The difference signal quantized at the quantization means
503
is fed to the variable length coder
504
to output as a bit stream and is also fed to the inverse quantization means
506
. Thereafter, a reconstructing process is performed to the coded difference signal. This reconstructing process, which is performed through the inverse quantization means
506
and the inverse DCT means
507
, is similar to the reconstructing process that is performed to the above described P-picture and B-picture. Namely, the reconstructed difference signal is added at the adder
508
with the predicted image signal previously given to the adder
508
to generate a reconstructed image signal.
The predicted image signal generated at the adder
508
is stored in the frame memory
509
as a reference image for use in a subsequent interframe prediction coding process. In the method represented by MPEG-2 (ISO/IEC-13818-2), the process of coding the input moving picture signal is therefore possible to reduce a redundancy in the spatial domain by means of the DCT transform as well as a redundancy in the time domain by means of the interframe prediction. Thus, moving pictures are coded efficiently. In addition, in the method represented by MPEG-2 (ISO/IEC-13818-2), the bit rate controller
505
may also be provided between the variable length coder
504
and the quantization means
503
. The bit rate controller
505
acquires information of the generated code quantity from the variable length coder
504
, determines a quantization scale so as to satisfy the restriction of the bit rate, and sends the quantization scale data to the quantization means
503
, thereby possibly controlling the generated code quantity.
Prior art for performing the coding process of MPEG-2 (ISO/IEC-13818-2) is specifically described below. A first example of the prior art is a method of MPEG-2 Test Model (Test Model 5, ISO/ICE JTC1/SC21/WG11/N0400, April 1993), which is known as a bit rate control method in a coding system with such a quantization process as described above. This system also employs a constant bit rate coding system, wherein the code quantity is made constant per certain unit time. For example, a known system controls a code quantity using a unit of group of pictures (GOP) that is a group of a plurality of pictures to be coded. The system adjusts a quantization scale to be set per macro block that is obtained from dividing a picture into 16×16 pixels in such a manner that a code quantity generated by the coding process on each GOP becomes constant.
The first example of the prior art employs the constant bit rate coding system. Therefore, in such a scene that a larger quantity of code is to be generated, it reduces the generated code quantity by increasing the quantization scale. To the contrary, in such a scene that a smaller quantity of code i
Boudreau Leo
Dang Duy M.
NEC Corporation
Whitham Curtis & Christofferson, PC
LandOfFree
Method and apparatus for coding moving picture at variable... does not yet have a rating. At this time, there are no reviews or comments for this patent.
If you have personal experience with Method and apparatus for coding moving picture at variable..., we encourage you to share that experience with our LandOfFree.com community. Your opinion is very important and Method and apparatus for coding moving picture at variable... will most certainly appreciate the feedback.
Profile ID: LFUS-PAI-O-3235276