A Proposed Model for Denoising Breast Mammogram Images

Abstract

Digital mammogram has emerged as the most popular screening technique to detect the breast cancer in its beginning and other abnormalities in human breast tissue. The paper’s objective is to list the types of noise obtained with the mammography process. Then, the most used filters used by the already proposed work in the breast cancer detection are introduced and then applied to denoise the images in the MiniMIAS database mammograms. The wiener, mean, median, adaptive median and Gaussian filters are applied on mammograms of different classes to be compared by measuring the mean squared error, the peak signal to noise ratio and the blurriness metrics. It is proved by the experimental results that the adaptive median filter is the best one denoise the salt and pepper noise. The mean filter is the best to denoise the Gaussian, speckle and poisson noise but with high blurriness value. This problem is overcome using the wiener filter in case of the Gaussian noise and using the median filter in case of the speckle and the poisson noise. The best filters for each type of noise are used together to remove any type of noise that may exist in the mammogram to achieve 11.3106 as the average mean squared error across the 322 mammograms and 37.9023 as the average peak signal to noise ratio which is the best combination leadings to the lowest mean squared error.