Classificação de texturas utilizando métodos de busca com informação

Abstract

Texture classification in images is a research subject of interest to many areas of application such as: face recognition, remote sensing and recognition of plant leaves. Several methods have been proposed in the literature. Recently, a new method for feature extraction, named Shortest Paths in Graphs (SPG), was presented. That method, in a simplified fashion, transforms an image into a graph and extracts some statistic measures based on Dijkstra’s minimum path through a set of points. In our work, we modify SPG in order to reduce the number of states visited in the graph and lower the computational cost. We carried out a number of experiments on the following image databases: Brodatz, UIUC, 2D HeLa and Virus. The best results achieved a reduction of 89.