httpiscsnthuedutwcourse20xxspringisa530100slidesl(參考版)
2025-07-24 23:33本頁面
【正文】 Bite Rate (Codebook size =128, BR= bit per pixel) . Hu Chang STSIGA Lena F16 Toys Tiffany Barb Gold Average bite rage Bite Rate (Codebook size =256, BR= bit per pixel) . Hu Chang STSIGA Lena F16 Toys Tiffany Barb Gold Average bite rage Bite Rate (Codebook size =512, BR= bit per pixel) . Hu Chang STSIGA Lena F16 Toys Tiffany Barb Gold Average bite rage Conclusion ? Two efficient algorithms for encoding the index value for image VQ are presented – Similarity between neighbor block – Associated index value – Use some different associated index table generation scheme to get better bit rate. ? Our approach improve about 50% bite rate over standard VQ, and about 25% bite rate over related methods. ? The performance of the proposed method is significantly affected by the index associated list. Thus, how to design a good IAI is the important issue in the future. 。 % better than conventional VQ Simple Tree Structure for Index Grouping Algorithm (STSIGA) 23 23 58 58 76 Record “ C=76 when L=23” Encoded block Unencoded block Encoding block C L STSIGA Step 1. Find the index value of encoding block by determining the closest index value of codeword in codebook Step 2. Compare the index value C of the encoding block with the index value L of the left bock and the index value U of the upward block. If C = L, then classify it as type 1, goto step 4. Otherwise, if C=U, then classify it as type 2, goto step 4. Step 3. Remove the index value of the encoded block in step 2, and encode the remaining blocks by Huffman coding. Classify these blocks as type 3. Step 4. Encode according to the type classified in step 2 and step 3. STSIGA Architecture of STSIGA Type 1 : 0 Type 2 : 10 Type 3 : 11 + Huffman code 0 0 1 1 STSIGA experimental results The bite rate of the proposed STSIGA is evaluated as follows. BR=[Con(1)x(1)+Con(2)x(2)+Con(3)x(2+p)] / (MxN) Con(i) = the number of the index value in type i, 1= i =3 ， p = the number of bits used in Huffman coding。 % better than conventional VQ % better than 、 % better than Hu – Chang’ method 。 % better than conventional VQ % better than 、 % better than Hu – Chang’ method 。 % better than conventional VQ How to generate the index associated list(IAL)? 0 1 2 . . . n1 n2 . . . 0 1 2 … n2 n1 P0~Pn1 denote the occurrence probability when the index value is different from the index value of left and upward blocks Codebook (n: codebook size) Index associated list P0 P1 P2 … Pn2 Pn1 ? Method C ISAIAL experimental results(三 ) Bite Rate (Codebook size =128, BR= bit per pixel) . Hu Chang Method C Lena F16 Toys Tiffany Barb Gold Average bite rate Bite Rate (Codebook size =256, BR= bit per pixel)
公司管理相關(guān)推薦
鄂ICP備17016276號-1