Overview

This dissertation, Joint image/video inpainting for error concealment in video coding by Liyong, Chen, 陳黎勇, was obtained from The University of Hong Kong (Pokfulam, Hong Kong) and is being sold pursuant to Creative Commons: Attribution 3.0 Hong Kong License. The content of this dissertation has not been altered in any way. We have altered the formatting in order to facilitate the ease of printing and reading of the dissertation. All rights not granted by the above license are retained by the author. Abstract: Abstract of the thesis entitled Joint Image/Video inpainting for Error Concealment in Video Coding submitted by Chen Liyong for the degree of Master of Philosophy at the University of Hong Kong in December 2007 The increasing demand for video communications and other multimedia application is one of the driving forces for the continual expansion of modern transmission mediums such as Internet and wireless networks. One inherent problem with these communication systems is that information may be distorted or lost during transmission due to channel noise. To mitigate this problem, the lost data is usually estimated from the received data, which is generally known as the error concealment problem. The purpose of this thesis is to study the use of inpainting techniques for error concealment in video coding and other applications. Image/video inpainting was originally proposed in the vision community in recent years, which aims at filling the missing content encountered in image/video processing. Generally, there are two categories of inpainting techniques, which is the spatial inpainting methods propagating information from the current image, and the temporal methods propagating information from neighboring images. In this thesis, we develop a framework to perform both spatial and temporal inpainting methods to realize the merits of these approaches. Firstly, a new joint motion-image inpainting method is proposed for error concealment of video coding and other applications. The proposed method extends a recent and efficient framework of motion inpainting originally developed in the vision community for video stabilizations also applicable to the error concealment problem. Furthermore, by introducing a mode selection mechanism for each pixel, the proposed method combines the motion inpainting with a novel image inpainting method to realize the merits of the two methods. Experimental results show that the proposed algorithm provide improved performance over conventional methods for the videos tested and it serves as an alternative to existing methods for video error concealment. Secondly, an improved joint image/motion inpainting method based on Bayesian motion estimation with occlusion detection is proposed. As the performance of motion inpainting depends highly on the accuracy of the estimated motion field, it will suffer from large occlusion areas where the motion field cannot easily be estimated. The proposed algorithm embeds spatial image inpainting into the framework of motion estimation, which gives an initial guess for the missing area and leads to an improved motion estimation result. Moreover, with the help of symmetric occlusion model, the occlusion area is identified and filled by spatial image inpainting. It also leads to a more reliable spatial-temporal mode selection criterion based on occlusion detection. Simulation results on real and synthetic images demonstrate the effectiveness of the proposed algorithm in inpainting the missing area. _______________________ An abstract of exactly 408 words DOI: 10.5353/th_b3955891 Subjects: Video compressionMultimedia communications

Full Product Details

9781361469033

Table of Contents

Reviews

Author Information

Tab Content 6

Countries Available