Pulse or digital communications – Bandwidth reduction or expansion
Reexamination Certificate
1999-10-08
2001-04-10
Kelley, Chris (Department: 2713)
Pulse or digital communications
Bandwidth reduction or expansion
Reexamination Certificate
active
06215820
ABSTRACT:
FIELD OF THE INVENTION
The present invention relates in general to the processing of digitized pictures, and, more particularly, to a picture coder having an output data stream with a constant bit-rate (CBR) to satisfy transmission and recording system requirements.
BACKGROUND OF THE INVENTION
A precise control on the number of bits per Group-Of-Pictures (GOP) is very important when recording a coded video sequence on an appropriate support, e.g., a CD-ROM, according to the MPEG2 standard. Since the GOP are regularly spaced, this permits simple play back operations, such as fast-forward and fast-reverse.
Only few pictures of the sequence are visualized during these reproducing operations. In particular, only the intra picture (I) is decoded. The I picture is the only one allowing a random access because they are independently coded from the rest of the sequence. Therefore, the reproduction includes searching within the bit-stream and decoding the first frame (I) of each GOP. This search may take place by a jump approximately close to the start of the GOP, and then by reading the bit-stream looking for the start code of the picture. Alternatively, a limited number of filler bits are inserted to make exactly constant the number of bits per GOP. This eliminates searching in the bit-stream.
Another advantage is the simplification of picture editing. It is possible to take a small portion of film, modify it, record it and insert it with precision in the same bit-stream interval. Without this capability, it would be necessary to read and save the entire bit-stream to assign the necessary space to the modified sequence.
There are two common objectives of any CBR system for controlling the bit-rate include the following. First, the desired or target bit-rate is preliminarily fixed and the pictures of a certain complete detail are taken into consideration for determining the coding quality. Therefore, the appropriate parameters for regulating the coding is such to obtain the predefined bit-rate. Second, a local quality is maintained as uniform as possible through a control process within the sequence.
The ability to accurately control the number of bits produced by each frame has consequences also upon the quality of the coded sequence. When the control system assigns the target bit-rate for the frame to be coded, it also distributes the remaining bits in the GOP among the frames that are still to be coded. If the bits produced by the frame coding are greater in number than the predicted ones, the GOP will have less bits than previously planned. Thereby, the control system will have to reduce the bits at disposal of the successive frames, which consequently, penalizes their quality. These problems occur mainly when there are substantial changes of the picture statistics, and, therefore, in their complexity. This typically occurs upon changes of a scene. The two objectives mentioned above are then conflicting.
In the ensuing description, reference will be made to the following publications on these topics and their pertinent contents are recalled as needed:
[1] International Standard ISO/IEC 13818 (MPEG2), “Information Technology—Generic Coding Of Moving Pictures And Associated Audio”, March 1994.
[2] B. G. Haskell, A. Puri, A. N. Netravali, “Digital Video: An Introduction To MPEG2”, Chapman & Hall, ISBN 0-412-08411-2.
[3] Test Model Editing Committee, “MPEG2 Video Test Model 5”, ISO/IEC JTC1/SC29/WG11, April 1993.
[4] G. Keesman, I. Shah, R. Klein-Gunnewiek, “Bit-Rate Control For MPEG Encoders”, Signal Processing: Image Communication 6 (1995) pp. 545-560.
[5] W. Ding, B. Liu, “Rate Control Of MPEG Video Coding And Recording By Rate-Quantization Modeling”, IEEE Transactions on Circuits and Systems for Video Technology, vol. 6, no. 1, February 1996.
[6] S. Bilato, G. Calvagno, G. A. Mian, R. Rinaldo, “Accurate Bit-Rate And Quality Control For The MPEG2 Video Coder”, proceedings of ICIP-97 (International Conference on Image Processing), Santa Barbara, 1997.
The article by Keesman et al. describes the finding of a compromise between the two contrasting objectives by introducing a technique based on defining a Bit Usage Profile which allows prediction on how the bits should be distributed within a picture. Such a technique is also described by Mian et al.
In the majority of video coders there is a control signal which directly influences the regulation of the local quality of a picture. Such a signal will be symbolized by Qj, where Q stands for a quantizer. A high value of Qj corresponds to a low quality of the image (coarse quantization) while a moderate value corresponds to a higher viewing quality (fine quantization). The objective of keeping a uniform quality is then equivalent to having a substantially constant control signal Qj within the picture and possibly throughout the entire sequence of pictures.
According to the system or model described in the “MPEG2 Video Test Model 5” article, hereinafter also referred to as the TM5 model, and also according to other CBR models cited in the bibliography, such as in the Ding et al. article, for example, control of the bit-rate is typically achieved in three phases:
1) estimating the global target bit-rate (Global Target Bit Allocation), wherein a prediction of the number of bits required for each picture of the sequence is generated. The complexity of the picture and the state of filling of the output data buffer must be considered;
2) determining the reference value Qj[n] of the local quantization parameter (Local Control) for the n-th macroblock (MB
n
) that satisfies a previously estimated target bit-rate;
3) calculating the effective quantization parameter mquant [n] for each macroblock, and generating it as a function of the filling state of the output buffer and of the activity of the macroblock itself (Adaptive Quantization).
According to the reference model disclosed in the TM5 article, the latter function is considered a substantial part of the bit-rate control. According to other known systems as disclosed in the Keesman et al. article, the Ding et al. article and the Mian et al. article, as well as in the context of the system of the present invention, this function is considered a separate problem when concerning the conversion of the control signal Qi [n] in the regulating parameters of the mquant [n] encoder. This conversion is according to the formula:
mquant
[
n
]
Qj
[
n
]
·
(
2
·
act
[
n
]
+
avg
—
act
act
[
n
]
+
2
·
avg
—
act
)
The variable act [n] is a measure of the detail of the picture based on the variance of the n-th macroblock, while avg_act is the average value of act [n] on the whole picture. The expression within the brackets is often referred to as the normalized activity Nact[f]. Therefore, in the ensuing description, the Adaptive Quantization will not be taken into consideration because it is a separate function, and it is not directly involved in the algorithm of the present invention.
Both the Global Target estimation and the Local Control use information on how to distribute the bits among the different pictures of the sequence and within the single pictures. This information may be derived from the previously coded pictures according to the TM
5
article, or by modeling the rate quantization curve according to the system described in Ding et al., or derived from a pre-analysis carried out on each picture, as in Keesman et al. and Mian et al.
In the system described by Mian et al., a pre-analysis is carried out to calculate the entropy of the DCT coefficients after having approximated the probability density function using histograms. By establishing a certain quantizer Q*, the entropy HN of the n-th quantizer DCT coefficient is calculated as follows. The variable Pk,n indicates the probability to obtain the K-th symbol at the output of the quantization. Therefore, the average entropy represents the actual bit-rate
Bagni Daniele
De Bei Mattia
Mian Gian Antonio
Sacchi Maria Luisa
Allen Dyer Doppelt Milbrath & Gilchrist, P.A.
Galanthay Theodore E.
Kelley Chris
STMicroelectronics S.r.l.
LandOfFree
Constant bit-rate control in a video coder by way of... does not yet have a rating. At this time, there are no reviews or comments for this patent.
If you have personal experience with Constant bit-rate control in a video coder by way of..., we encourage you to share that experience with our LandOfFree.com community. Your opinion is very important and Constant bit-rate control in a video coder by way of... will most certainly appreciate the feedback.
Profile ID: LFUS-PAI-O-2549890