Classification-based spatial error concealment for visual communications

Abstract

In an error-prone transmission environment, error concealment isan effective technique to reconstruct the damaged visual content.Due to large variations of image characteristics, differentconcealment approaches are necessary to accommodate the differentnature of the lost image content. In this paper, we address thisissue and propose using classification to integrate thestate-of-the-art error concealment techniques. The proposedapproach takes advantage of multiple concealment algorithms andadaptively selects the suitable algorithm for each damaged imagearea. With growing awareness that the design of sender andreceiver systems should be jointly considered for efficient andreliable multimedia communications, we proposed a set ofclassification-based block concealment schemes, includingreceiver-side classification, sender-side attachment, andsender-side embedding. Our experimental results provide extensiveperformance comparisons and demonstrate that the proposedclassification-based error concealment approaches outperform theconventional approaches. © 2006 Hindawi Publishing Corporation. All rights reserved.