Tumor Demarcation of Mammography Images Using Entropy with Different Window Sizes

Abstract

A technique is proposed for detection of tumor in digital mammography. We here proposed a statistical parameters such as probability and entropy based image segmentation of mammographic images. This algorithm is consists of three stages. First we compute probability of each quantization level of the image and replace every pixel by its probability to generate probability image. On this image we perform histogram equalization. Then we compute entropy for every pixel using different window sizes. By applying different thresholds on probability and entropy images a segmentation is performed.The results are compared with well known co-occurrence matrix method using probability and entropy. Experimental results show that our algorithm gives better results for mammographic images.