Classification-based motion adaptive de-interlacing algorithm

Liu Xiaodong,Ding Yong,Yan Xiaolang

DOI: https://doi.org/10.19650/j.cnki.cjsi.2011.10.018

2011-01-01

Abstract:To achieve high performance conversion from interlaced scan format to progressive scan format,a novel classification-based motion adaptive de-interlacing algorithm is proposed.The method mainly consists of five-field motion detection and classification,spatial-temporal weight calculation,Wiener-median filtering and 3-tap median filtering.In this method,the pixels to be inserted are firstly classified into different motion levels using the five-field motion detection and then corresponding algorithms are adaptively selected for de-interlacing,which effectively avoids the interpolation errors and improves the quality of the reconstructed image.Experimental results show that the objective performance of the peak signal to noise ratio obtained using this algorithm is 2.2 dB higher than the average results of existing algorithms;also the perception of subjective evaluation is satisfied.