WAVELET BASED FILTERS FOR ENHANCING DIGITAL MAMMOGRAMS

Abstract

The purpose of this study was to investigate the effectiveness of five wavelet–based filters in enhancing mammograms. Wavelet-based image enhancement was implemented by processing the Discrete Wavelet Transform detail coefficients. In addition, the wavelet-based filters were comparatively evaluated against five conventional Histogram Equalization Filters. Histogram equalization enhancement was based on modifying the image contrast by adjusting the gray-level probability density function (uniform, exponential, rayleight and two hyperbolic). These filters were applied to 130 digitized mammograms. The processed mammograms were blind-reviewed by an expert radiologist by means of eleven image quality parameters, including definition of masses, vessels and micro-calcifications. The wavelet-based filters enhanced significantly 6 of the evaluated parameters. On the other hand the two histogram equalization hyperbolic filters were found to effectively improve 7 parameters. The rest histogram equalization filters showed no significant image enhancement. Important results were the improved visualization of micro-calcifications and improvement of mammograms with fatty and fatty-granular content. Overall processing time was less than 3s for all filters on a typical desktop PC, rendering the application plausible for clinical routine