Loading...
Please wait, while we are loading the content...
Similar Documents
General method for eliminating redundant computations in video coding
| Content Provider | Semantic Scholar |
|---|---|
| Copyright Year | 2004 |
| Abstract | was coded using the conventional method. The complexity of each scheme was roughly estimated and expressed as a percentage of the complexity of the conventional scheme. Results obtained by processing the ‘Carphone’ video test sequence, with h = 60 and a quantisation step size of 20 for DCT coding, are shown in Table 1. The DVE algorithm is -20% more accurate than the LOG approach. The inaccuracy of the DVE algorithm is < 20%, and this drops to < 10% when the block size is reduced to 4 x 4. We now compare the average peak signal-to-noise ratio (PSNR) performance of the schemes for a given average bit rate. The performance results for two test sequences are shown in Table 2. The results show that the DVE algorithm has a performance -0.5dB better than that of the BM method and 0.35dB better than that of the LOG approach, and < 0.1 dB poorer than that of the EXH method. We have also included results arising from the use of a generalised form of Ramchandran’s algorithm [5] for DCT coding. The modest gains provided by the generalised algorithm (GTA) suggest that DVE obtains most of the gain available for solving eqn. 1. Carphone Foreman Rate PSNR Rate PSNR |
| File Format | PDF HTM / HTML |
| Alternate Webpage(s) | http://www.inesc-id.pt/ficheiros/publicacoes/558.pdf |
| Language | English |
| Access Restriction | Open |
| Content Type | Text |
| Resource Type | Article |
