High Performance DDDT-CWT Based Compressed Sensing Recovery of Images via Structured Sparsity

Due to its low encoding complexity, compressed sensing (CS) has gained wide attention in image processing related areas such as image compression, medical imaging and remote sensing. In existing research on CS based image processing, the commonly used sparse representation scheme for image recovery is the discrete wavelet transform (DWT), which is limited by poor directionality and lack of phase space information. What’s more, the structural information of transform-domain coefficients other than pure sparsity is seldom explored. In this paper, to improve the image recovery performance, we propose a new recovery method by adopting the double-density dual-tree complex wavelet transform (DDDT-CWT) as the sparse representation scheme. In addition, the structural characteristics of the DDDT-CWT coefficients are utilized as extra prior knowledge in the recovery process to further improve the recovery quality. Extensive simulation results have been conducted, and the results show that under the same compression ratio, the proposed method has achieved considerable PSNR gain compared with the traditional recovery algorithm.