Distributed video coding with particle filtering for correlation tracking

Abstract

One of the main challenges of distributed video coding (DVC) is that correlation among source and side information needs to be estimated and well modelled a priori. Since in DVC correlation dynamically changes with the scene, in order to get the full benefit from powerful distributed source code (DSC) designs, predicting and tracking correlation is essential. This paper proposes an adaptive scheme based on integrated particle filtering within the LDPC-based DSC that dynamically tracks the changes in image correlation to enhance belief propagation LDPC decoding. The system maintains its low encoder complexity, showing significant performance improvement compared to the case without dynamic particle filtering tracking.