NSCT-NLmeans based CS reconstruction for noisy image

Compressed sensing based on sparsity has been concerned in the field of signal reconstruction. Meanwhile nonsubsampled contourlet transform, proposed as a development to contourlet, not only provides flexible multi-scale, multi-direction sparse image decomposition, but also features with shift-invariance property which is beneficial to image denoising. This paper combines threshold operatoion in nonsubsampled contourlet domain with non local means filter for image denoising in the compressed sensing framework. The experiment results show that NSCT-NLmeans based algorithm outperformes the other multi-resolution and multi-directional transforms in recovering and denoising image simultaneously.