A Penalty-Resource Optimization Framework for Huffman-Treeased Variable Length Codes Decoding
Date Issued
2008
Date
2008
Author(s)
Wang, Sung-Wen
Abstract
Due to its sequential nature, speeding up Variable Length Decoding(VLD) is not as straightforward as to speed up the other modules of a decoding process by issuing several independent instructions simultaneously. The conservativeLD approaches restricted in a fixed algorithmic realization which might not be suitable for different applications’ need. In this dissertation, we design anlgorithmic level VLD optimization framework on the basis of penalty and resource, which can be automatically synthesized to match the resource constraints of target applications. More specifically, this work demonstrates anptimization methodology for minimizing the number of memory access subject to a memory size constraint for any Huffman-based variable length code decoding.e also propose a structure of memory-efficient hierarchical lookup table which can adapt to arbitrary-side growing Huffman-trees adopted in current CODECs. The proposed design decomposes theVLDinto a combination of multiple table lookups and imperative operations. A Viterbi-like algorithm is also proposed to efficiently find the optimal hierarchical table. More importantly,he Viterbi-like algorithm obtains the same results as that of the brute-force search algorithm. Simulation results show that the proposed hierarchical tableutperforms previous methods. To provide explicit illustration of hierarchical tables we also sketch the optimal solution for Huffman-trees adopted in currentODECs.
Subjects
Huffman Decoding
Variable Length Codes Decoding
Type
thesis
File(s)![Thumbnail Image]()
Loading...
Name
ntu-97-D92922017-1.pdf
Size
23.32 KB
Format
Adobe PDF
Checksum
(MD5):e0d86cd15b7e677001127abe54b62bf6