Huffman coding technique leads to loss-less coding and is used extensively in practical video compression systems. This technique assigns variable word length codes to different symbols based on the probabilities of the symbols. The unequal code word length of tlie symbols makes it difficult to implement high-speed Huffman decoders using pipelining and parallel processing. Previous approaches to pipeline the Huffman decoder algorithm lead to large increase in hardware complexity. In cost-effective decoders, we need to implement high-speed decoders using little or no hardware overhead. To this end, this paper proposes a new concurrent loss-less coder which requires no hardware overhead. Our loss-less coder imposes the code word length multiplicity constraint, i. e., it forces the code word lengths to be multiples of the speedup factor. This loss-less coder leads to slight degradation in coding performance (as measured in terms of the average code word length). We improve the performance of the concurrent coder by using conditional coding. These combined approaches lead to coding performance comparable with unconditional Huffman decoder and leads to about 10 times reduction in hardware compared with previous approaches (for the same speed-up).
|Original language||English (US)|
|Title of host publication||Workshop on VLSI Signal Processing 1992|
|Editors||Wojtek Przytula, Kung Yao, Rajeev Jain, Jan Rabaey|
|Publisher||Institute of Electrical and Electronics Engineers Inc.|
|Number of pages||10|
|ISBN (Electronic)||0780308115, 9780780308114|
|State||Published - 1992|
|Event||6th IEEE Workshop on VLSI Signal Processing - Los Angeles, United States|
Duration: Oct 28 1992 → Oct 30 1992
|Name||Workshop on VLSI Signal Processing 1992|
|Conference||6th IEEE Workshop on VLSI Signal Processing|
|Period||10/28/92 → 10/30/92|
Bibliographical noteFunding Information:
'This research was supported by the Office of Naval Research under contract number NO001 4-91-J - 1008
© 1992 IEEE.