Pulse or digital communications – Bandwidth reduction or expansion – Television or motion video signal
Reexamination Certificate
1999-07-06
2002-08-06
Britton, Howard (Department: 2613)
Pulse or digital communications
Bandwidth reduction or expansion
Television or motion video signal
C375S240120, C708S203000
Reexamination Certificate
active
06430222
ABSTRACT:
BACKGROUND OF THE INVENTION
(1) Field of the Invention
The present invention relates to a moving picture coding apparatus which performs an orthogonal transform of an input picture by dividing it into two-dimensional blocks and quantizes and codes the data, and more particularly relates to a moving picture coding apparatus which quantizes and codes data of picture by extracting a specific area from the frame of picture in a manner so as to enhance the picture quality of the extracted area.
(2) Description of the Related Art
Recently, visual communications services such as video telephone, video conference etc., show great prospect as an efficient way of using ISDN (Integrated Services Digital Network) and GSTN (General Switched Telephone Network) services. Therefore, investigations into high efficiency coding techniques aiming at efficient transmission of such moving pictures have been intensively carried out. Investigations of these techniques were based on statistic properties of pictures to remove redundancy contained in the pictures to thereby reduce the amount of information. As the coding techniques, the hybrid coding technique is well known which is a combination of motion-compensated prediction and the discrete cosine transform.
However, in the case of transmission at a low bit rate, the quality of the regenerated picture is degraded due to noise generation. Hence, improvement against this has been desired. As a method for improvement of picture quality, a technique has been investigated in which specific areas are extracted from the input picture so that the quantization stepsize is controlled separately for each area. As an example of this, there is an idea for improving the picture quality subjectively, by extracting the face area from the input picture so that areas excepting the face area hereinafter referred to as ‘background’ will be quantized based on a quantization stepsize greater than that for the face area, to thereby reduce the amount of coding allotted for these areas while a greater amount of coding will be allotted for the face area (for example, R. H. J. M. Plompen, et al.: “An image knowledge based video codec for low bitrates”, SPIE Vol. 804 Advanced in image processing (1987)).
Referring now to
FIG. 1
, a conventional example using motion compensated prediction and a two-dimensional orthogonal transform will be explained. In
FIG. 1
, the conventional system includes: a frame memory section
81
; a subtracter
82
; an orthogonal transformer
83
; a quantizer
84
; a coding section
85
; a buffer memory section
86
; an inverse quantizer
87
; an inverse orthogonal transformer
88
; an adder
89
; a frame memory section
90
; a motion detecting section
91
; a motion compensation predicting section
92
; an area extracting section
93
; and a coding control section
94
.
Now, suppose a picture is input to frame memory section
81
. The input picture is a digitized picture obtained from a video camera or the like, and is stored in frame memory section
81
and is divided into blocks of N×M pixels (N and M are natural numbers). The picture stored in frame memory section
90
and the picture stored in frame memory section
81
are input to motion detecting section
91
where the motion vectors are detected. In motion compensation predicting section
92
, motion compensated prediction values are determined based on the motion vectors and the picture stored in frame memory section
90
. The power of the differences between the motion compensated prediction values and the input picture and the power of the input picture are compared to choose interframe prediction or intraframe prediction.
In subtracter
82
, the difference between the input picture in frame memory section
81
and motion compensated prediction value from motion compensated predicting section
92
is calculated for each block unit. Orthogonal transformer
83
implements the two-dimensional orthogonal transform of the pixels in each block and outputs the transform coefficients to quantizer
84
. Quantizer
84
quantizes the transform coefficients based on the quantization stepsize output from coding control section
94
. Coding section
85
implements entropy coding of the quantized output from quantizer
84
to generate coded information. Buffer memory section
86
accumulates the coded information in order to make the system capacity match the transmission rate of the network.
The output from quantizer
84
is also input to inverse quantizer
87
, where it is inverse quantized to produce transform coefficients. In inverse orthogonal transformer
88
, the transform coefficients are subjected to the two-dimensional inverse orthogonal transform, and the result and the motion compensated prediction value from motion compensation predicting section
92
are added in adder
89
. The thus added picture is stored in frames memory section
90
to be used in the next process. Coding controlling section
94
receives valid/invalid information about the face area and the background area as the output from area extracting section
93
and the occupied amount of buffer memory section
86
, and determines the quantization stepsize based on these pieces of information. For example, the valid information indicating the face area and the invalid information indicating the background area are output from area extracting section
93
to coding controlling section
94
, which chooses a quantization stepsize for the face area smaller than that for the background area, based on the quantization stepsize determined from the occupied amount of buffer memory section
86
.
The above technique only sets two levels of quantization stepsize for the face area and the background area, and only specifies the fact that the quantization stepsize for the face area is smaller than that for the background area. Therefore, for actual application of this configuration to a real moving picture coding apparatus, the following operation may and should be done, for instance. Coding controlling section
94
defines dQf, dQb, which are the amounts of displacement for the extracted result from the area extracting section for the quantization stepsize Q which is determined based on the occupied amount of the coded information in buffer memory section
86
. Quantizer
84
quantizes the data for the face area using Q−dQf and the data for the background using Q+dQb.
However, in this method, noise etc. existing in the background will also be quantized even though the quantization stepsize is set large, an increased amount of coding arises, which means a reduction in the amount of coding which can be allotted to the face area. Further, since a picture greatly differs from the previous frame when a scene change occurs, if the amount of coding for the background is low, the picture quality of the background area will not become stabilized at an early stage and the reproduced picture becomes rough and ugly, tending to cause block distortion and mosquito noise around the face area due to sharp increase in amount of coding.
Since only the occupied amount of the coded information in buffer memory
86
is referred to when coding controlling section
94
performs control, the quantization stepsize is determined for that buffer's occupied amount, which makes it difficult to attain the desired, proper, amount of coding. Consequently, variations in picture quality become large between frames, lowering the picture quality.
Moreover, though frames after a scene change, in general, have strong correlation, it is difficult to control quantization so as to attain the target, proper, amount of coding since this system does not use the information from the quantization stepsize of the previous frame or the information as to the amount of coding.
When the amount of displacement of Q (dQf) is kept fixed independent of the size of the face area, the amount of coding per frame may become too high or too low with respect to the target amount of coding depending upon the size of the face area, resultantly the picture quality of the reproduc
Britton Howard
Nicon & Vanderhye, .P. C.
Sharp Kabushiki Kaisha
LandOfFree
Moving picture coding apparatus 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 coding apparatus, we encourage you to share that experience with our LandOfFree.com community. Your opinion is very important and Moving picture coding apparatus will most certainly appreciate the feedback.
Profile ID: LFUS-PAI-O-2973277