Segmentação de imagens via competição entre regiões fuzzy: abordagens paramétricas e não-paramétricas

Abstract

Image segmentation is an important step in the existing pattern recognition and feature extraction systems. The primal goal is to subdivide an image into its constituent regions. Among the many works published in the literature, we highlight the methods based on variational principles, which are effective tools to formulate stable techniques for image segmentation. Here, we studied variational models for image segmentation present in literature, in which will be discussed their mathematical formulations and characteristics of their minimization processes. Among these techniques will be explored in detail the method Fuzzy Region Competition and segmentation models that can be derived from its energy functional. Inspired by this method, three modifications were proposed based on these models that minimize some of their limitations and which emerges as alternatives for real world image segmentation. The proposed modifications were validated by the good results obtained using natural images, textured and noisy. In addition, some comparative results are shown with other similar techniques with the objective of reporting the effectiveness of the proposed models.