Orthogonality Deficiency Compensation for Improved Frequency Selective Image Extrapolation

TL;DR

This paper introduces an efficient image extrapolation algorithm that uses orthogonality deficiency compensation to improve signal modeling, resulting in better PSNR and visual quality in image and video applications.

Contribution

It proposes a novel orthogonality deficiency compensation method to enhance basis function projection in image extrapolation algorithms.

Findings

Improves PSNR by up to 2 dB.

Achieves better visual quality in block loss concealment.

Effective for both structured and smooth image areas.

Abstract

This paper describes a very efficient algorithm for image signal extrapolation. It can be used for various applications in image and video communication, e.g. the concealment of data corrupted by transmission errors or prediction in video coding. The extrapolation is performed on a limited number of known samples and extends the signal beyond these samples. Therefore the signal from the known samples is iteratively projected onto different basis functions in order to generate a model of the signal. As the basis functions are not orthogonal with respect to the area of the known samples we propose a new extension, the orthogonality deficiency compensation, to cope with the non-orthogonality. Using this extension, very good extrapolation results for structured as well as for smooth areas are achievable. This algorithm improves PSNR up to 2 dB and gives a better visual quality for concealment of block losses compared to extrapolation algorithms existent so far.