Segmentation and Characterization of Breast Tumor in Mammograms

Abstract

The objective of this paper is to present a novel approach for the segmentation and characterization of breast tumor. The authors have developed an efficient tumor segmentation technique using conventional Ostu thresholding method aided by morphological reconstruction filtering technique. The parameters extracted are area, major and minor axis length, eccentricity, solidity, average gray level, standard deviation, and entropy. This method is simple, fast and versatile. This can be applied for the segmentation of all types of breast tumor irrespective of their size. The proposed technique has been implemented on MATLAB 7.3 and results are compared with the existing techniques such as standard Ostu thresholding method and region growing method proposed by Zucker. The simulated result shows that the performance of the proposed technique is better than that of existing techniques.