A High-Throughput VLSI Architecture Design Of Canonical Huffman Encoder

In this brief, a high-throughput Huffman encoder VLSI architecture based on the Canonical Huffman method is proposed to improve the encoding throughput and decrease the encoding time required by the Huffman code word table construction process. We proposed parallel computing architectures for frequency-statistical sorting and code-size computational sorting. This architecture results in a process of building a tree and assigning symbols that can be completed by scanning the data only once. This solves the problem of the low efficiency of the traditional algorithm, which needs to scan the data twice. Consequently, in addition to the advantages of the high compression ratio inherited from the Canonical Huffman, the proposed architecture has overridden advantages for a high parallelism processing capacity. The experimental results showed that the proposed architecture decreased the encoding time by 26.30% compared to the available Huffman encoder using the standard algorithm when encoding 256 8-bit symbols. Furthermore, the VLSI architecture could further decrease the encoding time when encoding more 8-bit symbols. In particular, when encoding 212,642 8-bit symbols, the proposed VLSI architecture could reduce the encoding time by 87.40%. Thus, compared with the traditional Huffman encoders, this brief achieved the improvement of coding efficiency.