A Hybrid Image Denoising Method
Hari Om, Mantosh Biswas · International Journal of Computer Applications · 2012
In this paper, a hybrid image denoising method that is based on locally adaptive window-based maximum likelihood (LAWML) and NeighShrink.The LAWML is doubly stochastic process models which denoise an image by exploiting the dependency of local wavelet coefficients within each scale.The LAWML needs a global optimal neighboring window.The NeighShrink thresholding scheme uses the immediate neighboring coefficients based on block thresholding.It uses a suboptimal universal threshold and identical neighbouring window size in all wavelet subbands.The NeighShrink and LAWML always produce an over-smoothed image like the Weiner filter in which many of the detail coefficients are lost during threshold evaluation.This proposed method overcomes these disadvantages and, as a result, it provides significant improvement in visual quality i.e.Peak-to-Signal Noise Ratio (PSNR) of a noisy image.