Image Denoising Using Hybrid Contourlet and Bandelet Transforms

This paper proposed a new multiscale and multidirectional image representation method named CBlet transform. It combines the new contourlet transform with the second-generation bandeletzation procedure. Thereinto, the contourlet transform captures image discontinuous points and links them into linear structures, then the bandeletization procedure pursuits the linear structures adoptively and further removes their correlation. The CBlet transform obtains much sparser representation than the new contourlet transform at the same redundancy. Numerical experiments on image denoising show that the proposed CBlet transform can outperform the new contourlet transform both in term of PSNR and in visual quality.