Combination of color-based segmentation, Markov random fields and multilayer perceptron

Abstract

Angioectasias are lesions characterized by specific features, related to their color and shape. Since the high prevalence of angioectasias in the small bowel, it is of great importance the development of a method to correctly localize these lesions within the intestinal tissue. Since the differences found in the color of the lesions, when compared with other lesions and the normal tissue, it was developed a method based of the probability segmentation of pixels, with a Markov Random Field property to improve the neighborhood of the lesion. This was done with the CIELab color space, since it was found that has high efficiency in differentiating colors in an image.