Error concealment for classified texture images
Jaroslav Polec, Michal Pohancenik, Sandra Ondrušová, Kvetoslava Kotuliaková, T. Karlubikova · 2009
For error masking in the static images, it is good to know image classification. Having an effective classificator, we should achieve selecting a method good enough for restoration of missing image segments. It is very important especially for texture images. Common methods usually fail in these cases. It is necessary to select the method, which is dedicated for synthesizing an image or for inpainting. In this paper, three methods of extrapolation and texture synthesis are shown. They are analysed in terms of suitability for restoration of missing texture image segments. For one type of texture classification, the analysis of several categories is made. As the missing image segments, block of 8 times 8, macroblock of 16 times 16 and strip of macroblocks are considered. The evaluation is done using objective as well as subjective criteria.