Denoising of Image Using Least Minimum Mean Square Error
Vinay Kumar Dawar, Mohit Bansal · 2012
In this paper, image denoising by linear minimum mean square-error estimation (LMMSE) scheme is proposed and also the determination of best suited wavelet for image denoising has been discussed. The over complete wavelet expansion (OWE) in noise reduction is used for taking the effective result instead of orthogonal wavelet transform. A vector has been designed by the combining the pixels at the same spatial location across scale to explore the strong inter-scale dependencies of OWE and apply LMMSE to the vector. Now, the performance evaluation of the proposed scheme is done by using different wavelet family. To measure the denoising performance, two criteria are used, first is signal information extraction and second is distribution error criterion. The best suite wavelet, which achieves best results between these two criteria, can be selected from wavelet family. To exploits the wavelet intrascale dependency and image discrimination, estimate the wavelet coefficients statistics and wavelet coefficient is classified by Context modelling.