Intra coding with one-dimensional data block for AVS video
Wang Qiju · Computer Engineering and Applications Journal · 2013
The fixed 3-tap filter in AVS intra prediction cannot adjust to various textures well, resulting in low coding efficiency of intra coding. Considering that texture consistency of adjacent samples is much stronger, adaptive filtering prediction based on one-dimensional block is proposed. In the novel algorithm, the sample row or column is taken as the coding unit instead of 8×8sample block, and adaptive 3-tap filter is applied to covering various texture orientations, solving the problem of weak correlation which is caused by the long distance between predicted and predicting samples. To avoid the transmitting cost of adaptive filter coefficients, the adjacent coded one-dimensional sample block is taken as the target block and its adjacent block as the source to generate prediction, and via least square method, it's easy to obtain adaptive filter coefficients which are directly facilitated to the coding of current one-dimensional sample block. Experimental results show that the novel intra coding method outperforms conventional intra coding in AVS video by an average of 0.324 dB with maximum of 0.879 dB. The adaptive intra filtering coding based on one-dimensional block can get a stable gain of average 0.120 dB compared with intra coding based on template matching prediction. The novel intra coding method demonstrates a good potential in future AVS 2.0 standard.