Multiresolution joint bilateral filtering with modified adaptive shrinkage for image denoising

This paper proposes an effective image denoising of gray level images in wavelet transform domain using joint bilateral filter and modified adaptive shrinkage. The input image is first decomposed using 2D-discrete wavelet transform. The multiresolution joint bilateral filtering is then applied to the approximation sub-band of the decomposed image. Flat and edge regions are obtained using the α-map computed from wavelet transform coefficients of LH, HL, and HH bands. Noise is removed in the flat regions by Inner Product method. After removing noise in the flat regions, further noise removal is done in the edge regions using NeighShrink SURE shrinkage functions. The modified wavelet coefficients in the flat and edge regions are combined and filtered by using Gaussian low pass filter. Finally the denoised output image is reconstructed using 2D-inverse discrete wavelet transform. Experimentation has been carried out on set of standard test images using the proposed algorithm and its performance is evaluated and compared with existing state of art methods using PSNR, EKI and Computation time. Experimental results show that the proposed algorithm can effectively reduce noise without losing sharp details in the noisy images and is suitable for commercial low-cost imaging systems.