An efficient motion adaptive deinterlacing algorithm using improved edge-based line average interpolation

Abstract

In this paper we present an efficient motion adaptive de-interlacing algorithm. The motion detection method and the intra-field interpolation are two main factors that affect the efficiency of motion adaptive de-interlacing. To improve the accuracy of motion detection, five consecutive interlaced fields are used and we also introduce a Gaussian filter to eliminate the affection of noise. The adaptive threshold adjusting is also helpful for the determination about which region is moving. For the moving area, we also introduce a texture detector to distinguish texture area and smooth area. Our algorithm adopts conventional edge-based line averaging (ELA) interpolation in the texture area and modified ELA interpolation is adopted in the smooth area. As a result, our algorithm outperforms other conventional non-motion compensated methods both objectively and subjectively.