| 摘要 |
<p>A method and apparatus that distinguishes the static component of a frame from its dynamic component and provides a computationally efficient method for background extraction. If a block in a current frame is determined to correspond to a static component, it is not encoded or sent. A preferred embodiment of the present invention works on the block level and has two basic parts. First, the difference in pixel intensities between each block in the current frame and its best-matched corresponding block in a previous frame are determined. These differences are then processed further to reduce the effects of noise in subsequent calculations before they are passed to the second part. The second part applies a concept called herein 'center of gravity' to determine a degree of similarity between two blocks in the respective current and previous frames. The center of gravity is determined for a chosen block (in the current frame) and compared with its best match (in the previous frame). A large discrepancy in the two figures indicates dis-similarity and a small discrepancy indicates similarity.</p> |
