A Voting-based Intra Deinterlacing Method for Directional Error Correction

This paper presents a voting-based deinterlacing method for directional error correction (VDD), which can reduce wrong edge direction in conventional schemes. The VDD consists of four steps and their corresponding interpolation methods: a Filtered Modified-Edge-based Line Averaging (FM-ELA), a Majority Voting-based Interpolation (MVI), a Majority voting-based Direction Averaging method (MDA), and the Direction-Oriented Interpolation (DOI). Each step can preserve edge directions and enhance visual quality in high spatial frequency regions through the analysis of local region features. After each missing pixel is categorized into one of three groups, a different deinterlacing technique is applied in order to provide the best result. Although it has a little bit higher complexity than previous methods, VDD can preserve delicate edges compared with previous methods. Experimental results show the improved subjective and objective quality of the reconstructed images.