Classification of Mammogram Images Using Discrete Wavelet Transformations

A fractional part of biggest wavelet coefficient is enough to describe the characteristics of an image texture. Based on this concept a mammogram classification algorithm is developed. Using this classification algorithm, we classified the mammogram image into different classes as normal, benign and malignant. Ten percent of the images in each class are used for creating a class core vector. This class core vector acts as the base for the classification. The Euclidean distance is measured between the test image feature vector and the class core vector of the each class. A test image is classified into the appropriate class, which has minimum Euclidean distance measured between the test image and class core vector. Using this classification algorithm we classified 134 mammogram images into the exact class out 162 test images in the dataset. This algorithm results a detection rate of 75% for normal images, 88 % of detection rate for malignant and 100% detection rate for benign images respectively. The overall detection rate is 83%.