A low complexity alf based on inter-channel correlation between chroma and luma in HEVC

The adaptive loop filter (ALF) is a Wiener algorithm based technology which is applied to the reconstructed frame to improve the image quality in high-efficiency video coding (HEVC). In the conventional ALF algorithm, filter coefficients are re-designed for 16 times to obtain the best sets in one frame and also to increase the external memory access and computational cost significantly. This contribution proposes an inter-channel transformation scheme based on the correlation between chroma and luma to alleviate both computing time and memory bandwidth. First, the proposed algorithm generates the cross-correlation matrix and auto-correlation matrix in chroma samples which only spends one-quarter computational costs due to the resolution difference. Then, it applies the inter-channel transformation to derive the filter coefficients. The proposed methods not only eliminate a large number of multiplication operations, but also reduce the run-time by 58.70% averagely during whole ALF encoding process, while a negligible loss 0.074 BD-PSNR in the Y component.