An improved image denoising technique using cycle spinning
Sayed Mohammad Ebrahim Sahraeian, Farrokh Marvasti · 2007
Denoising of corrupted images has been a classical problem in image processing. In this paper we propose a new approach for image noise reduction using wavelet transform. In this method an improved version of thresholding neural networks (TNN) is used to find the optimum threshold values in the sense of minimum mean square error (MMSE). Based on these optimum thresholds a novel cycle-spinning based method is used to reduce image artifacts. In this method, we utilize two thresholding schemes as the thresholding operator of cycle-spinning. A neighbor dependent thresholding scheme is employed as its first shrinkage step and a simple wavelet thresholding with the optimum derived threshold values is used as the second thresholding step. Using this approach we will achieve a smooth, artifact free denoised image. Experimental results indicate that the proposed method outperforms several other established wavelet denoising techniques, in terms of peak-signal-to-noise- ratio (PSNR) and visual quality.