Detection method and removing filter for corner outliers in highly compressed video

Abstract

We propose a detection method and a removal filter for corner outliers in order to improve visual quality in highly compressed video. Corner outliers are detected using the direction of edge going through a block-corner and the properties of blocks around the block-corner. The proposed filter for removing corner outliers, which compensates the stair-shaped discontinuities around edges using the adjacent pixels, is applied to the detected area. Simulation results show that the proposed method improves, particularly in combination with deblocking filters, the visual quality remarkably.