Comparison of texture based algorithms for the detection of masses in digitised mammograms

A comparison of methods to detect masses in digitised mammograms based on textural information is investigated. Masses in mammograms are usually "smoother" with lower textural content than the surrounding tissue. Our algorithms enhance and segment the image based on texture. We implemented four algorithms: 1) hexagonal wavelets, 2) Laws texture energy masks 3) wavelet enhanced Laws, and 4) steerable Alters enhanced Laws. We tested each of these algorithms on a standard database, and varied parameters showing interesting properties in each one.