MOVING-PICTURE TEMPORAL SCALABLE
CODING METHOD, CODING APPARATUS,
DECODING METHOD, DECODING
APPARATUS, AND COMPUTER PROGRAM
BACKGROUND OF THE INVENTION
The present invention relates to a moving-picture temporal scalable coding method and a moving-picture temporal seal- 10 able coding apparatus, a moving-picture temporal scalable decoding method and a moving-picture temporal scalable decoding apparatus, and also a computer program for performing the coding or the decoding method.
Moving-picture coding is classified into simple one-layer 15 coding and scalable coding for encoding two-layer bitstreams. The latter allows decoding a bitstream of a base layer only and also decoding a bitstream of an enhancement layer, decoded base-layer and enhancement-layer pictures being combined to reproduce high-quality pictures. 20
Scalable coding is classified into SNR (Signal-to-Noise Ratio), spatial, and temporal scalable coding. The temporal scalable coding is to decimate, for example, a 60-fps (field per second) interlaced image per field to obtain a 30-fps image and encode this 30-fps image while predicting the remaining 25 non-encoded fields by using a locally decoded image of the encoded fields and encode prediction residuals.
In known moving-picture temporal scalable coding, a 60-fps interlaced moving-picture video signal is divided into even-number fields and odd-number fields. 30
The even-number fields are subjected to coding while the odd-number fields are subjected to delay.
In coding, a video signal carrying 30-fps even-number fields is coded into a bitstream and quantization resultants (not a bitstream but signal components at least quantized). 35 The coding technique may be MPEG inter-picture predictive coding or intrafield coding.
The quantization resultants are subjected to local decoding to be reproduced into a local decoded picture. The local picture is subjected to inter-picture prediction to produce a 40 predictive signal for each odd-number field.
In delaying, each odd-number field is delayed until the predictive signal is produced based on each even-number field, as explained above.
The predictive signal is subtracted from an odd-number- 45 field delayed signal to obtain a prediction residual.
The prediction residual is subjected to DCT (Discrete Cosine Transform). The resultant 8x8 DCT coefficients are subjected to quantization at a given step width. The resultant fixed-length coefficients (prediction residual) are subjected to 50 variable-length coding to obtain a bitstream.
This bitstream is multiplexed with the bitstream already obtained from the even-number fields, as an output movingpicture bitstream under temporal scalable coding.
In summary, under the known temporal scalable coding, an 55 interlaced moving-picture video signal is divided into evennumber fields and odd-number fields. The even-number fields are converted into a base-layer bitstream while the odd-number fields an enhancement-layer bitstream, or vice versa.
The base-layer bitstream and the enhancement-layer bit- 60 stream are multiplexed with each other to form an output moving-picture bitstream under temporal scalable coding, as illustrated in FIG. 1.
In FIG. 1, a sign "field" indicates one field of an interlaced video. The numbers attached to the signs "field" indicate the 65 order of coded pictures. Base-layer pictures come before enhancement-layer pictures for bi-directional prediction of
the enhancement-layer pictures, even though the former pictures come after the latter pictures in the time domain. The reverse order is further required among the base-layer pictures when bi-directional prediction is performed for these pictures.
In known moving-picture temporal scalable decoding, a moving-picture bitstream obtained from a 60-fps interlaced moving-picture video signal by temporal scalable coding, is divided into a base-layer bitstream, an enhancement-layer bitstream, and a scale factor.
The base-layer bitstream is decoded so that a 30-fps video signal is reproduced. The reproduced signal carries evennumber fields of the 60-fps interlaced moving-picture video signal. The reproduced signal is subjected to inter-picture prediction to produce a prediction signal for odd-number fields of the interlaced moving-picture video signal.
The enhancement-layer bitstream is subjected to variablelength decoding so that variable-length codes of prediction residual is reconverted into fixed-length codes.
The fixed-length codes are subjected to dequantization at a given quantization parameter to be reproduced into DCT coefficients of prediction residual.
The DCT coefficients are subjected to inverse DCT so that 8x8 DCT coefficients are converted into a decoded prediction-residual signal.
The decoded prediction-residual signal is added to the prediction signal already produced to form a 30-fps decoded video signal. This decoded signal carries the odd-number fields of the 60-fps interlaced moving-picture video signal.
The odd-number fields of the 30-fps decoded video signal and the even-number fields of the 30-fps video signal are selected in synchronism with the scale factor. The latter video signal carrying the even-number fields have already been decoded and delayed until the former video signal is decoded.
The odd-/even number field selection reproduces the 60-fps interlaced moving-picture video signal.
As explained, under the known temporal scalable coding, an interlaced moving-picture video signal is divided into even-number fields and odd-number fields. The even-number fields are converted into base-layer bitstream while the oddnumber fields an enhancement-layer bitstream, or vise versa.
The known temporal scalable coding, however, has several drawbacks.
Base-layer coding causes many prediction errors in motion-compensated inter-picture prediction due to many aliasing components involved in field pictures.
Enhancement-layer coding suffers inaccurate inter-picture prediction due to difference in parity (even/odd) of fields between pictures to be coded and prediction reference pictures.
These two factors drastically lower coding efficiency in the known temporal scalable coding compared to other coding techniques.
SUMMARY OF THE INVENTION
A purpose of the present invention is to provide a movingpicture temporal scalable coding method and a moving-picture temporal scalable coding apparatus that achieve high coding efficiency in coding of interlaced moving-picture video signals, a moving-picture temporal scalable decoding method and a moving-picture temporal scalable decoding apparatus for decoding the video signals coded by the coding method and apparatus, respectively, and also a computer program for performing the coding or the decoding method.
The present invention provides a temporal scalable moving-picture video signal coding method comprising the steps