Edge information based effective intra mode decision algorithm
Yule Yuan, Xiaohang Sun · 2012
In this paper, we present very simple, yet efficient, algorithms for intra mode decision algorithms in AVS video coding based on edges detection and neural network. The technique uses the SOBEL operator to check the edge information from the total image before its intra prediction. Then the prediction mode of each 8×8 block can be decided based on its edges. In our paper, we design two schemes to decide the intra mode. In our first method, the intra mode is determined by compare the total number the feature points in each sub-blocks. It can save the encoding time of AVS part 2 between 30 to 40%. It is a very fast method, but the bit stream is correspondingly increased. To reduce the gain of the bit rate, we design another effective method employing neural network classifier which trained by the edge points to decide the mode of the block for AVS intra prediction. Because only one prediction mode is chosen for RDO calculation using the proposed algorithm, simulation results also show that the proposed scheme achieves up to 9% computational saving with no video quality degradation, compared with results of the existing method.