A reconstruction algorithm based on 3D tree-structure Bayesian compressive sensing for underwater videos

Abstract

The traditional underwater video coding has relative high demand of Underwater Acoustic Channel and the underwater video scenes are complex and instable. To deal with these problems, this paper presents a reconstruction algorithm based on three dimension (3D) tree-structure Bayesian compressive sensing for underwater videos. On the encoder side, analog color coded aperture compressive temporal imaging system (CACTI) encodes the video signal. On the decoder end, based on the model of Bayesian compressive sensing, by exploiting the 3D tree structure of the wavelet and Discrete Cosine Transformation(DCT) coefficients, a Bayesian compressive sensing inverse transform algorithm is derived to reconstruct color video frames from single-channel compression measurements. The experimental results show that the algorithm is able to reconstruct complex video scenes more accurately.