An In-Loop Filter Based On Human Visual System

High Efficiency Video Coding (HEVC) is under development by the Joint video coding standardization project of International Telecommunication Union Telecommunication Standardization Sector (ITU-T) Video Coding Experts Group (VCEG) and International Organization for Standardization (ISO)/Interaational Electrotechnical Commission (IEC) Moving Picture Experts Group (MPEG). The current in-loop filter has two tools, deblocking filter (DF) and sample adaptive offset (SAO). The aim of deblocking filter is to decrease the blocking artifacts introduced by block-based prediction and quantization. SAO is aim to reduce the ringing artifacts, which mainly stems from quantization errors of transform coefficients. SAO including two techniques, band offset (BO) and edge offset (EO). In this paper, two algorithms are proposed to improve the performance of deblocking filter and sample adaptive offset respectively.To improve the accuracy, the deblocking filter is modified with the parameter information of intra coded blocks in this paper. Considering that the masking effect of human visual system (HVS) will eliminate the visibility of the blocking artifacts, a combined luminance and directional-texture (CLDT) masking effect model-based DF is proposed. The experimental results show that the proposed filter decreases the bit rate by0.86%on average against high efficiency video coding test model with the same objective video quality.The SAO band offset (BO) is perfect only the constructed samples in a Coding Tree Block (CTB) spread over four continuous bands, but not adaptive if the signal focuses on a range narrower or wider than four bands. Two additional types are proposed in this paper to solve this problem. If the signal is compacted to two discontinuous bands, another BO type is added. The proposed method can significantly reduce the bit rate compared with High Efficiency Video Coding test model, introducing-0.4%,-2.5%coding gains for luma, chroma on average.