Multi-Hypothesis Compressed Video Sensing Technique

TL;DR

This paper introduces a novel multi-hypothesis compressive sensing method for video that achieves superior recovery performance with faster computation, combining sparsity and Tikhonov regularization.

Contribution

It proposes a new iterative algorithm integrating multi-hypothesis, sparsity, and Tikhonov regularization for efficient video reconstruction in compressive sensing.

Findings

Outperforms Elasticnet in recovery accuracy

Faster than Elasticnet, comparable to Tikhonov in speed

Validated by extensive simulations

Abstract

In this paper, we present a compressive sampling and Multi-Hypothesis (MH) reconstruction strategy for video sequences which has a rather simple encoder, while the decoding system is not that complex. We introduce a convex cost function that incorporates the MH technique with the sparsity constraint and the Tikhonov regularization. Consequently, we derive a new iterative algorithm based on these criteria. This algorithm surpasses its counterparts (Elasticnet and Tikhonov) in the recovery performance. Besides it is computationally much faster than the Elasticnet and comparable to the Tikhonov. Our extensive simulation results confirm these claims.