An Improved Color Image Denoising using Spatially Adaptive Statistical Dependency in Undecimated Wavelet Transform Coefficients

Abstract

In color image denoising, it is very important to preserve the edge and color information of the images. The standard discrete wavelet transforms provide a means of implementing a multiscale analysis, based on a critically sampled filter bank with perfect reconstruction, but shift-variance and directional selectivity is not supported by them which are necessary for color image processing. This is due to the decimation used in traditional wavelet transform and can be covered by undecimated wavelet transform. An adaptive color image denoising algorithm using spatial correlation in undecimated wavelet transform is proposed in this paper that preserves the color image features more efficiently. The noisy color image is first decomposed into different subbands of frequency and orientation responses using the undecimated wavelet transform. Then the spatial correlation of wavelet coefficients in a noisy color image is achieved by the coefficients multiplication of adjacent decomposition levels. The threshold value is calculated from the spatially correlated wavelet coefficients by computing the noise standard deviation from the correlated diagonal subband of first and second decomposition level. Finally, all noisy spatially correlated wavelet coefficients are relatively well denoised by soft thresholding. The experimental results demonstrate that the proposed color image denoising algorithm preserve the edges and color components very well while reducing noise. The efficiency and performance of this algorithm is compared with state-of-art denoising algorithms based on Peak Signal to Noise Ratio and visual perception.