Background Preserved and Feature-Oriented Contrast Improvement Using Weighted Cumulative Distribution Function for Digital Mammograms

Abstract

Digital mammography is an inevitable source for the early detection of breast cancer. The limitations of this imaging modality tend to impede the contrast and brightness of digital mammograms, which may adversely affect the accuracy of breast cancer diagnosis. Many of the existing contrast enhancement methods pose the challenges of over-transformation of background, noise amplification, and lack of details preservation. The proposed technique named Background Preserved and Feature-Oriented Contrast Improvement (BPFO-CI) using Weighted Cumulative Distribution Function (WCF) for digital mammograms aims to solve these limitations of contrast enhancement techniques. The BPFO-CI uses a threshold to control the background-over-transformation, and further, it divides the remaining dynamic gray levels of the foreground into two symmetric regions. The first region is partitioned by its mean gray level, and the second region is partitioned by its median, generating four distinct ranges of gray levels. The algorithm further computes the weighted cumulative distribution function using a local intensity adjustment factor which is the weight of each range that controls over-transformation of each partitioned range of intensities. Finally, the original intensities of the input mammogram are mapped with the computed gray levels. The performance measures of BPFO-CI were recorded as Mean Brightness Error (MBE), Structural SIMilarity index (SSIM), and Peak Signal Noise Ratio (PSNR) by experimenting on 322 mini-MIAS mammograms. The quantitative measures of this proposed technique are confirmed to be better than the existing methods. The qualitative measure, namely human visual perception also endorses its merit.