Moving-picture coding and decoding method and apparatus with reduced computational cost
View Patent ↗A coded moving-picture signal is decoded by a resolution-converting motion compensation process and a resolution-converting inverse discrete cosine transform, both of which decrease the resolution of the picture, thereby reducing the amount of reference picture data that has to be stored and accessed. The reference picture data may also be stored in a compressed form. The resolution conversion and compression processes may also be used in the coding of the moving-picture signal. The resolution-converting inverse discrete cosine transform may be performed by output of intermediate results that have not been combined by addition and subtraction in a butterfly computation.
1. A moving-picture decoder for decoding a predictively coded moving-picture signal, comprising:
a predictive decoding unit decoding said predictively coded moving-picture signal with reference to a reference picture signal, thereby obtaining an output picture signal;
a data compressor coupled to said predictive decoding unit, compressively coding said output picture signal, thereby obtaining a compressed picture signal;
a memory coupled to said data compressor, storing said compressed picture signal; and
a data decompressor coupled to said memory, expanding said compressed picture signal, thereby obtaining said reference picture signal,
wherein said data compressor comprises:
a block partitioning unit dividing said output picture signal into data blocks;
a transform processor independently transforming each of said data blocks; and
a coding unit coding each of the transformed data blocks,
wherein said data decompressor comprises:
a decoding unit decoding each of the coded transformed data blocks, thereby obtaining decoded transformed data blocks;
an inverse transform processor performing an inverse transform on each of the decoded transformed data blocks to obtain decoded data blocks; and
a block assembler reassembling the decoded data blocks,
wherein said decoding unit performs a dequantizing process that multiplies all data values in said decoded transformed data blocks by a predetermined power of two, then adds an integer bias value centering said data values according to data probability distributions in each quantizing interval.
2. A method of decoding a predictively coded moving-picture signal, comprising the steps of:
(a) decoding said predictively coded moving-picture signal with reference to a reference picture signal, thereby obtaining an output picture signal;
(b) compressively coding said output picture signal, thereby obtaining a compressed picture signal;
(c) storing said compressed picture signal in a memory; and
(d) expanding said compressed picture signal, thereby obtaining said reference picture signal,
wherein said step (b) further comprises the steps of:
(e) dividing said output picture signal into data blocks;
(f) independently transforming each of said data blocks; and
(g) coding each of the transformed data blocks,
wherein said step (d) further comprises the steps of:
(h) decoding each of the coded transfonned data blocks, thereby obtaining decoded transformed data blocks;
(i) performing an inverse transform on each of the decoded transformed data blocks to obtain decoded data blocks; and
(j) reassembling the decoded data blocks,
wherein said step (h) further comprises the steps of:
performing a dequantizing process that multiplies all data values in said decoded transformed data blocks by a predetermined power of two; and
adding an integer bias value centering said data values according to data probability distributions in each quantizing interval.