Direct mapping architecture for JPEG Huffman decoder

摘要

A novel decoding algorithm for the JPEG Huffman code is presented, in which the alternating current (AC) Huffman table is partitioned into four modules. According to this partition, a direct mapping technique is developed to decode Huffman codewords without any searching or matching operations. Therefore, an algorithm based on this technique can save several controlling circuits and will be simple for the hardware implementation. There are also two other merits: first, the run/size symbols and codeword lengths can be obtained directly; and, next, it is not necessary to store any AC Huffman codewords in the memory space at the decoding end. As a result, the overall behaviour of the developed algorithm is better than the conventional methods. This is illustrated through a mathematical verification using the Big-O notation and a hardware simulation using hardware description language. Tables of comparison results are included in the paper. Furthermore, a VLSI architecture of this proposed algorithm is also developed and included.