Motion compensation as sparsity-aware decoding in compressive video streaming

Abstract

We consider a video transmission system where the transmitter performs merely direct compressive sensing with no other forms of encoding/processing and the burden of quality video sequence reconstruction falls solely on the receiver side. We show that effective implicit motion compensation can be carried out at the receiver/decoder via iterative sparsity-aware recovery on adaptively forward-backward estimated Karhunen-Loève bases. Experiments illustrate these developments.