Adaptive edge preserving image denoising in wavelet domain

Abstract

With the advancement in internet and multimedia technologies a huge amount of multimedia data in the form of audio video and images has been used in many areas like medical fields satellite data digital forensics surveillance systems etc This has led to higher demand for multimedia data like images with high visual quality However digital images can be corrupted by noise during the process of acquisition and transmission degrading their quality So some image denoising techniques must be needed to restore the quality of image A major challenge for a denoising algorithm is to improve the visual appearance of an image while preserving relevant features such as edges during the denoising process Wavelet transforms have been widely used for edgepreserving image denoising due to their ability of decorrelation separation of noise and useful signal Numerous denoising approaches based on wavelet transforms have been proposed in last few decades and the research is still continuing in this direction Researchers all over the world are trying to improve the performance of wavelet based denoising approaches The work presented in this thesis is the next step in this series A number of problems in existing wavelet based denoising methods are identified and their solutions are proposed Generally wavelet transforms are not adaptive local structures of image are not considered during decomposition An adaptive Haar wavelet transform tetrolet transform was used to resolve this issue but this tetrolet system was originally nonredundant in nature while for image denoising redundant information is helpful In addition conventional thresholding schemes which used a global threshold for filtering the noise in wavelet coefficients can also remove high frequency structures such as edges Also the noise variance which is used in the computation of threshold is usually kept fixed through all resolution scales However the noise strength decreases with the increase in resolution scale A denoising technique using redundant tetrolet transform