Contourlet based mammographic image enhancement

摘要

In original mammographic images obtained by X-ray radiography, only a small part of detected information is displayed to the human observer. A method aimed at minimizing image noise while optimizing contrast of mammographic image features is presented in this paper, for more accurate detection of microcalcification clusters. The method is based on the contourlet transform, which is a multiresolution, local and directional image representation. The difference from wavelet and other multiscale expansion lies in that the contourlet transform is constructed by using non-separable filter banks in discrete-domain, thus it can effectively capture important features of images. The enhancement procedure consists of two steps: noise filtering by the Stein's thresholding and denoised contourlet coefficients modification via a nonlinear mapping function. The experimental results have shown an improved visualization of significant mammographic features by the proposed method. A comparison with other enhancement algorithms is also discussed by employing a measure named target to background contrast ratio using variance.