Análise de Imagens Através de Segmentação Não-Supervisionada de Texturas

Unsupervised texture-based image segmentation is a important step in many computer vision algorithms, retrieving regions that may be identified as objects or be subject of further segmentation. However, there isn't any known technique with optimal response to every kind of natural image. Each segmentation algorithm has its strengths and weakness, affecting the image partition quality. In this work, the study of three mainstream texture segmentation techniques is shown. The algorithms were applied to three class of natural images and the results compared, regarding either segmentation quality and time efficiency.