On the adaptive motion estimation in video coding based on video content analysis

摘要

In this paper, we propose a novel, efficient and fast motion estimation (ME) algorithm for High Efficiency Video Coding (HEVC). The proposed work is based on video content that gives us the provision to select a best macro block mode and suitable search pattern for a given slice of the video frame. Each frame is divided into multiple slices based on a tiered scene labelling mechanism and a suitable search pattern is estimated for each slice by investigating texture of the slice. The interspace distance between a macro block in current frame to the candidate macro block in the previous frame is calculated for all the macro blocks in a slice, to find the block movement factor (BMF) which identifies a appropriate search pattern for that slice. Also, the disparity between the neighbouring blocks is measured for each slice to predict the mode to be considered for block matching in next frame. The experimental analysis is performed with x265 video codec, which shows considerable amount (20%) of improvement in compression performance.