Application of Discrete Multi-wavelet Transform in Denoising of Mammographic Images

Abstract

Objectives: Clinical and surgical procedures heavily depend on acquiring and transmitting medical information by electronic devices and media in the form of medical images. In this process the quality of these images are degraded by noise, making denoising mandatory. Methods/Statistical Analysis: In this paper de noising algorithms in wavelet domain for mammographic images (used for detection of breast cancer in women) are considered. A modified approach for de noising of mammographic images using Multi Wavelet Transform has been proposed with four different thresholding techniques.These Multi wavelets possess important properties like orthogonality, symmetry and compact support simultaneously.Performance of denoising methods improves considerably due to these properties. Findings: Proposed method is applied on large data set of digital mammographic images freely available as mammographic data base by MAIS. DMWT based denoising scheme with four different types of thresholding estimates, namely Bayesian shrink, Visu shrink, neighborhood shrink and modified neighborhood shrink is applied on data set and compared the performance of proposed denoising algorithm with dwt based existing denoising methods. Higher the value of σ (noise variance) for Gaussian noise considered lower is the value of PSNR, obtained, whatever is the transformation and techniques applied. Even among the DMWT: GHM,CL and SA4, SA4 gives the best response for all the cases. Applications/Improvements: The results are compared on the basis of Peak Signal to Noise Ratio (PSNR) in dB. Results clearly indicate the superiority of the proposed method in all the four cases over exiting wavelet based method. This new improved method will help a lot in diagnosis of breast cancer more accurately and the correct prognosis can save many lives.