Abstract
It is known that mammogram screening is aimed to detect non-homogeneous and subtle symptoms of breast cancer. These features are rarely visible owing to marginal visual thresholds between the specific abnormality and the complex background tissues. Computer aided detection and diagnosis techniques (CAD) for breast cancer are reliant upon the degree of improvement in contrast and sharpness (of the tumour region), provided by a mammogram enhancement approach. UM based enhancement model yields better perception results making it feasible for processing mammographic images. This also ensures coherence with Human Visual System (HVS) characteristics but with certain associated challenges. This paper aims to narrate a comparative evaluation of various UM based approaches (in context to enhancement of mammographic images) on the basis of visual analysis as well as objective evaluation using standard Image Quality Assessment (IQA) metrics. In this paper, 10 UM based enhancement approaches are evaluated starting from the traditional Linear UM (LUM) along with subsequent evolutions (since the past two decades) covering the recent Non-Linear UM. Finally, an improved HVS based UM approach using Non-Linear Polynomial Filters (NPF) has been discussed as a robust solution to provide enhancement of mammograms with different nature of background tissues as well as types of masses. The outcomes of the study suggested that non-linear UM approaches are more suited towards enhancing the mammographic mass (tumour) region with respect to its background. (C) 2017 Elsevier B.V. All rights reserved.