Inpainting Strategies for Reconstruction of Missing Data in Images and Videos: Techniques, Algorithms and Quality Assessment

Abstract

This paper describes a method and algorithm for spatially-temporally consistent image and video completion. We propose modification of an image inpainting algorithm based texture and structure reconstruction. Proposed method allows to remove static and dynamic objects and restore missing regions using spatial and temporal information from neighboring frames. This paper also focuses on a machine learning approach for no-reference visual quality assessment for image and video inpainting. Experimental comparisons to state-of-the-art inpainting methods demonstrate the effectiveness of the proposed approaches.