| 摘要 |
An encoder executes an iterative search method for selecting deblocking parameters for coded video. According to the method, a decoded picture may be deblocked according to parameters associated with a multi-dimensional deblocking vector and an error, called the "deblocking error" may be estimated therefrom. If the estimated error exceeds a predetermined threshold, then alternate deblocking vectors may be created, each advanced from the current deblocking vector in a respective dimension. The method may deblock the decoded picture according to each advanced vector and may estimate errors from each of the deblocking of each advanced vector. Finally, the deblocking vector may be revised for a next iteration according to a gradient derived from the estimated deblocking errors of the vector dimensions. This gradient-based search method may converge on a final set of deblocking parameters in an efficient manner. |
