Video Inpainting using self-adaptive GMM with Improved Inpainting Technique

Abstract

Abstract: Nowadays photography and videography have become part of life. There are many challenging tasks in videography, one of these is video inpainting. Repairing the damaged videos or removing and filling undesired objects in videos is defined as video inpainting. In this paper, an object-splitting video inpainting is proposed. The background subtraction is implemented by self-adaptive Gaussian Mixture Model (SAG). The moving foreground and static background are inpainted by using enhanced inpainting technique with improved patch priority calculation. The structure consistency patch matching is proposed to search for the similar patch in the source region to fill the target region. The video inpainting results are obtained for the own video. The proposed inpainting technique is implemented for state of art videos utilized in other related works. The experimental results show that the proposed technique attained impressible inpainted videos compared to related works.

Index Terms: Video inpainting; SAG, background subtraction; patch priority; Structure consistency