Summary: | 碩士 === 國立成功大學 === 電機工程學系碩博士班 === 98 === Most video compression standards, such as MPEG-2 and H.264/AVC, use variable length coding to remove statistical redundancy. Since the codewords are variable length, how to represent the codeword table to reduce the size of storage is quite important. However, the throughput of decoding may be suffered from codeword reduction. Besides, both MPEG-2 and H.264/AVC VLC decoder can not be combined directly due to the difference of codeword table of the two standards. In this thesis, we proposed a low-cost VLC decoder which supports MPEG-2 and H.264/AVC and achieves real-time decoding.
In this thesis, we first proposed a 2-D to 1-D tree translation scheme to minimize the number of stored bits. However, the throughput is also limited by the tree search decoding flow. Then we decomposed the VLC tree by use tree partition with sub-tree classification. Tree partition with sub-tree classification not only reduces the size of codeword storage but also improve the throughput compared to the proposed 2-D to 1-D tree translation scheme. The proposed method can support MPEG-2 and H.264/AVC decoding and has 20% of area reduced compared to the combination of H.264 decoder and MPEG-2 decoder. The throughput of the proposed design also meets real-timely video decoding.
|