Adaptive Gaussian Low-pass Pre-filtering for Perceptual Video Coding
Wahyu Wiratama, Yong-Jo Ahn, Ismail Marzuki, Donggyu Sim · IEIE Transactions on Smart Processing and Computing · 2018
This paper proposes adaptive Gaussian low-pass pre-filtering for perceptually highquality video encoding, considering the characteristics of human visual perception. The proposed pre-filtering adaptively removes high-frequency components that are not visually perceptible. In the proposed algorithm, a pixel-wise adaptive Gaussian filter is generated based on a local contrast threshold for a block that is not highly textured. Each pixel for the block is adaptively filtered, and the original pixel is replaced by the filtered value or a weighted one that is a mixture of the filtered and original values for edge preservation. The proposed algorithm can remove visually imperceptible components, and leads to bit reduction without visual deterioration of the coded videos. Experimental results show that the proposed algorithm can reduce the bitrate by up to 30% with minimal computational complexity and minimal visual quality loss for several video sequences.