Abstract:
The purpose of this work is to develop an algorithm for denoising images corrupted by various noise models. The noise degrades quality of the images and makes interpretations, analysis and segmentation of images harder. In this research work a novel image denoising paradigm is envisaged to work for various noise models. The image is decomposed by multilevel wavelet decomposition using Quadrature mirror filters and then thresholded. The decomposition level is determined by the image resolution. The property that image structural details remain present at each level whereas noise is partially eliminated within subbands, is being exploited. The lower resolution wavelet detail bands are interpolated to the original image size which partially recaptures the image pixels besides facilitating matrix multiplications. An innovative wavelet synthesis approach is conceived based on wavelet scale correlation of the concordant detail bands such that the reconstructed image fabricates a denoised image. An entropy reduction criterion is used in parallel to PSNR for analytical analysis of the results. The subjective analysis supported by the objective analysis reveals that the results image denoising through the proposed scheme are satisfactory in various noise environments. Discrete wavelet transform (DWT) using scale correlation is a denoising approach that removes the noise effectively than the simple wavelet decomposition. The detail coefficients in concordant bands are correlated and then synthesized after soft thresholding, which suppresses noise but signifies smooth intensity variations. The wavelet coefficients of noise have much trivial correlation than the wavelet coefficients of boundaries that propagate along the scale. Scale multiplication improves the localization accuracy significantly while keeping high detection efficiency. The combination of noise filtering coupled with boundary detection in a single algorithm produces the filtered version of the image.
