Abstract

Ranklets are non–parametric, multi–resolution and orientation selective features modelled on Haar wavelets. A ranklet–based image representation is proposed in this paper in order to solve a two–class classification problem. The first class is constituted by masses, breast tumors with size ranging from 3 mm to 30 mm, whereas the second class is constituted by non–masses. Masses and non–masses are both extracted from the University of South Florida (USF) mammographic image database, submitted to the ranklet transform and finally classified by means of a Support Vector Machine (SVM). Experiments demonstrate that the proposed image representation solves succesfully the two–class classification problem. Furthermore, it achieves an improvement over the pixel–based and wavelet–based representations tested on the same dataset by one of our previous works.