No-reference objective quality assessment and enhancement for compressed images and videos

Abstract

This thesis proposes new and effective methods to assess the objective quality of compressed digital images and videos, and to enhance the reconstruction quality of them using multiple compressed copies. Digital images and videos are shared over the Internet by millions of users everyday. Such contents are usually highly compressed for efficient transmission at the expense of fidelity. The end users may want to know the quality of these images and videos after lossy compression, should she wish to archive or further process them. There may also be several differently compressed copies of the same content residing at different locations on the web. The end user may like to compare the quality between them so as to pick the best one, or even generating a “super copy” with image quality superior to any of the copies available.