Image analysis – Image compression or coding
Reexamination Certificate
1999-12-20
2003-05-20
Do, Anh Hong (Department: 2624)
Image analysis
Image compression or coding
C375S240000, C386S349000
Reexamination Certificate
active
06567554
ABSTRACT:
BACKGROUND OF THE INVENTION
1. Field of the Invention
The present invention concerns a data coding method and an apparatus therefor which enable, for example, coded audio and video data to be stored respectively into units, as a row of packs that is to be reproduced within a prescribed period of time. Especially, the invention concerns a data coding method and an apparatus therefor which make it possible to realize, while keeping the quality of a picture image stably, encoding, before encoding of audio and video data, that permits depiction of navigation data such as data length and starting address that are calculated from a value corresponding to the amount of codes of the coded data.
2. Related Art
In recent years, a data compaction system for use on a moving picture image has hitherto been internationally standardized as an MPEG (Moving Picture Experts Group) system. This MPEG system is known as a system for performing variable compaction of video data. In the MPEG system, there are defined compaction systems which are called “MPEG 1” (MPEG phase 1) and “MPEG 2” (MPEG phase 2).
Concretely, the MPEG is prepared by several techniques being combined with one another. First, a time redundancy portion is reduced by subtracting a picture image signal that has been obtained by being decoded by a motion compensation unit from an input picture image signal.
As the method of prediction, there are three modes, as fundamental modes, i.e., a mode in which prediction is performed from past picture images, a mode in which prediction is performed from future picture images, and a mode in which prediction is performed from both past picture images and future picture images. Also, each of these modes can be used by being switched in units of a macroblock (MB) composed of 16 pixels×16 pixels. The method of prediction is determined according to the picture type (“Picture_Type”) that has been imparted to an input picture image. As the picture types, there are a one-directional between-picture prediction encoded picture image (P-picture), bi-directional between-picture prediction encoded picture image (B-picture), and intra-picture independently encoded picture image (I-picture). In the P-picture type (one-directional between-picture prediction encoded picture image), there are two modes one of which is to encode by performing prediction from past picture images and the other of which is to independently encode a macroblock without performing relevant prediction. In the B-picture (bi-directional between-picture prediction encoded picture image), there are four modes, a first one of which is to perform prediction from future picture images, a second one of which is to perform prediction from past picture images, a third one of which is to perform prediction from both past picture images and future picture images, and a fourth one of which is to encode independently without performing any prediction. In the I-picture (intra-picture independently encoded picture image), all macroblocks are each independently encoded. It is to be noted that the I-picture is called wan “intra-picture” and that, therefore, the one-directional between-picture prediction encoded picture image and the bi-directional between-picture prediction encoded picture image can each be referred to as “a non-intra-picture”.
In the motion compensation, by performing pattern matching of the movement regions in units of a macroblock, a motion vector is detected with a half pixel precision and prediction is made after shifting of the macroblock to an extent corresponding to the thus-detected motion vector. The motion vector includes horizontal and vertical motion vectors, and this motion vector is transmitted as additional messages for macroblock along with an MC (Motion Compensation) mode that indicates where prediction is made from.
The pictures from the I-picture to a picture that immediately precedes the next I-picture are called “GOP (Group Of Picture)”. In a case where pictures are used in accumulation media or the like, approximately 15 pictures or so are generally used as 1 GOP.
FIG. 1
illustrates a fundamental construction of a video encoder that is among audio/video encoding apparatuses, to which the MPEG is applied.
In this
FIG. 1
, an input picture image signal is supplied to an input terminal
101
. This input picture image signal is sent to a calculating unit
102
and to a motion compensation and prediction unit
111
.
In the calculating unit
102
, a difference between a picture image signal, which has been decoded in the motion compensation and prediction unit
111
, and the input picture image signal is determined, and a picture image signal corresponding to this difference is sent to a DCT unit
103
.
In the DCT unit
103
, the differential picture image signal that has been supplied is subjected to orthogonal transformation. Here, the DCT (Discrete Cosine Transform) means an orthogonal transformation through which an integrating transformation that uses a cosine function as an integrating kernel is changed to a discrete transformation that is made into a finite space. In the MPEG system, two-dimensional DCT is performed of 8×8 DCT blocks that have been obtained by dividing the macroblock into four parts. It is to be noted that in general a video signal is composed of a large amount of low frequency band components and a lesser amount of high frequency band components and that, therefore, when performing DCT, the coefficients thereof are concentratedly gathered into the low band. Data that has been obtained by performance of the DCT in the DCT unit
103
is sent to a quantizing unit
104
.
In the quantizing unit
104
, quantization is performed of the DCT coefficients from the DCT unit
103
. In the quantization performed in this quantizing unit
104
, a two-dimensional frequency of 8×8, which constitutes a quantizing matrix is weighted by visual characteristics. The value that has been resultantly obtained is further made scalar-fold by a quantizing scale. And using the thus-obtained value as a quantizing value, the DCT coefficient is divided by this value. It is to be noted that when performing inverse quantization, by a decoder (video decoder), of coded data after the encoding performed by this video encoder, multiplication of it is made by the quantizing value that was used in the video encoder. As a result of this, it is possible to obtain a value that is approximate to the original DCT coefficient. Data that has been obtained by the quantization made in the quantizing unit
104
is sent to a variable length coder (VLC)
105
.
The VLC
105
performs variable length coding on the quantized data from the quantizing unit
104
. In this VLC
105
, of the quantized values, with respect to direct current (DC) components, coding is performed using DPCM (differential pulse code modulation) that is one of the prediction coding techniques. On the other hand, with respect to alternating current (AC) components, so-called “Huffman coding” is performed in which so-called “zigzag scan” is performed from a low band to a high band and, by counting the run length and effective coefficient value of a zero as being one piece of significant event, a code having a shorter code length is allotted to the data sequentially from one, the probability of whose occurrence is higher. Also, to the VLC
105
there are also supplied from the motion compensation and prediction unit
111
motion vector and prediction mode messages, whereby the VLC
105
outputs these motion vector and prediction mode messages as well as the variable coded data as additional data with respect to the macroblock. Data that has been obtained by the variable length coding performed in the VLC
105
is sent to a buffer memory
106
.
The buffer memory
106
temporarily stores therein the variable length coded data from the VLC
105
. Thereafter, the coded data (the coded bit stream) that has been read out from the buffer memory
106
at a prescribed transfer rate is output from an output terminal
113
.
Also, the amount of codes generated in macro
Sugahara Takayuki
Suzuki Junzo
Berkowitz Marvin C.
Do Anh Hong
Nath Gary M.
Novick Harold L.
Victor Company of Japan , Limited
LandOfFree
Data coding method and apparatus therefor does not yet have a rating. At this time, there are no reviews or comments for this patent.
If you have personal experience with Data coding method and apparatus therefor, we encourage you to share that experience with our LandOfFree.com community. Your opinion is very important and Data coding method and apparatus therefor will most certainly appreciate the feedback.
Profile ID: LFUS-PAI-O-3043061