Taik

EFFICIENT HUFFMAN DECODING WITH TABLE LOOKUP Mohamed F. Mansour DSPS R&D Center, Texas Instruments Inc., USA mfmansour@ ti.com
ABSTRACT We describe an efficient algorithm for Huffman decoding using table lookup. The algorithm is optimized for ROMbased Huffman decoding. It is a two-step process of prefix template matching followed by a direct table access. We propose an efficient algorithm for choosing the prefix templates according to different optimization criteria. Also, we propose different implementations for the prefix template procedure. Index Terms— Huffman decoding, Table lookup 1. INTRODUCTION Huffman codes [1] have been widely used for source coding and have shown high efficiency in exploiting the source redundancy. Huffman codes along with run-length codes have been widely used in most international multimedia standards (e.g., MPEG and ISO standards [5], [7]). Huffman decoding can be implemented with a lookup-table (LUT) [2] or multiple lookup-tables [3]. If a single LUT is used, the decoder throughput can be one codeword per cycle whereas the throughput for multiple lookup tables is not deterministic and in the worst case it equals the number of lookup tables (assuming each LUT is processed in a single cycle). The single LUT approach is usually adopted in high efficiency Huffman decoder while the multiple LUTs approach is usually used in low-power systems. In this work, we propose a novel LUT-based approach for Huffman decoding. The decoder has an LUT for a set of prefix templates for the table codewords. Each prefix template is associated with a direct access table for the children codewords. During decoding, the input bits after the prefix template are used to directly address the associated codeword table to retrieve the correct codeword and its length. We propose a novel approach for designing the prefix templates which depends on a generic optimization criterion that can be adjusted to the system. We propose different criteria that can be employed...