LAUSR

The novel contribution of this work is to introduce a new promising method for the analysis of textures: the Decimal Descriptor Patterns (DDP). Two best known methods of texture measures, always considered as references in image analysis are chosen for comparison: the Local Binary Patterns (LBP) and the Grey Level Co-occurrence Matrix (GLCM). We realized numerous experimentations for analyzing the brain Magnetic Resonance (MR) images in order to demonstrate the interest of our proposition. We used the 3D Brainweb database with two brain MR images sequences, and different levels of noise and intensity non-uniformity. This way accuracy of the three methods is tested in front on the image artifacts. Tests of classification are performed in the same conditions of work by means of the classifier multiclass Support Vector Machines (SVM). Experimental results demonstrate clearly the robustness and the stability of the proposed approach with respect to the noise level, intensity non-uniformity and to different T1- and T2- weighted MR images.