Method and circuit arrangement for undersampling in the case of movement

Abstract

In the case of the storage or transmission of moving images, it is frequently necessary to reduce the large number of information items by means of suitable compression methods. In order to estimate the movement, the image to be coded is in this case split into blocks, and each of these reference blocks R is compared with blocks of the same size of the preceding image, and a movement information item is derived. Conventional methods for movement estimation in this case use all the pixels in the reference block R and in the search area S. Since this very complicated method requires very fast, and thus expensive, hardware, the so-called quincunx undersampling is applied to the search area S. In the case of the reference block R, one quincunx undersampled pixel is used for even parity of the movement vector and the other quincunx undersampled pixel is used for odd parity, depending on which position is being investigated. For the search area S, it is then necessary to store only half the pixels which are required for movement estimation. The memory requirement and the bandwidth of the memory are thus reduced by half.