Haar Transform and SVD Based Contrast Enhancement and Segmentation for Mammographic Images

Low contrast and poor quality are the main problems in the production of mammographic images. This thesis proposes an improvised method using the wavelet transform and Single value decomposition approach to enhance the image and then mean shift clustering to segment the image in order to clearly identify the edges of the tumor. In this proposed method a mammographic image is decomposed using wavelet transform. The wavelet frequency field, an image’s edge feature information and detail information are distributed in high-frequency sub-images. But there is still more detailed information in the other sub-image(s). In order to obtain more detailed image information, all high-frequency sub images are further decomposed using Haar transform. The noise in the high frequency field is reduced by the soft-threshold method. Singular value decomposition (SVD) is applied on LL Band for enhancement of low frequency co-efficient values and for high-frequency coefficients the enhancement is obtained by multiplying different weight values in different sub-images. Later, the enhanced image is obtained through the inverse wavelet transform and inverse Haar transform.
Finally, the image’s histogram is stretched by nonlinear histogram equalization. Towards, segmentation process we use the Moving K-mean clustering algorithm. Experiments showed that this method does not only enhance an image’s details but can also preserve its edge features effectively.