| 摘要 |
In the data-reducing encoding/decoding of digital image signals, in many cases a movement-compensated interpolation of pixel values is used as a basis. To determine the movement vectors for pixel blocks, a comparison is carried out for each pixel block from the current image within a search region which consists of an enlarged pixel block. By means of MSE (mean square error) or MAE (mean absolute error), within the search region, the block with the highest correlation is determined and the respective movement vector is determined from its spatial position in comparison with the block from the current image. This process is extremely computationally intensive and requires very high clock rates. Even in the case of hierarchical search processes, complex and fast hardware is still necessary. According to the invention, the movement estimation is carried out by evaluating only specific bits, for example the MSB (most significant bit) or both the MSBs, of the pixel values. For this purpose, these bits are logically linked by means of EXOR and/or EXNOR functions and the individual results for the various block positions of the current block within the search window are in each case combined. By means of a subsequent minimum or maximum comparison, the respective vector for the current block is then determined. <IMAGE> |
