Medical Images Segmentation

Abstract

This chapter presents a new and efficient unsupervised color segmentation scheme named GBOD to detect visual objects from medical color images and to extract their color and geometric features, in order to determine later the contours of the visual objects and to perform syntactic analysis. The presented method is a general-purpose segmentation algorithm, and it produces good results from two different perspectives: (a) from the perspective of perceptual grouping of regions from the natural images (standard RGB) and also (b) from the perspective of determining regions if the input images contain salient visual objects. We present a unified framework for image segmentation and contour extraction that uses a virtual hexagonal structure defined on the set of the image pixels. This proposed graph-based segmentation method is divided into two different steps: (a) a presegmentation step that produces a maximum spanning tree of the connected components of the triangular grid graph constructed on the hexagonal structure of the input image and (b) the final segmentation step that produces a minimum spanning tree of the connected components, representing the visual objects, by using dynamic weights based on the geometric features of the regions. The presegmentation step uses only color information extracted from the input image, whereas the final segmentation step uses both color and geometric and spatial configuration of image regions. Despite the majority of the segmentation methods, our method does not require any parameter to be chosen or tuned in order to produce a better segmentation, and thus our method is totally adaptive. The entire approach is fully unsupervised and does not need a priori information about the image scene. The new proposed algorithm is compared with other two well-known segmentation algorithms: the color set back-projection algorithm and the local variation algorithm. In order to evaluate these segmentation algorithms, we used error-measuring methods that quantify the consistency between them. The evaluation is made on a database with color medical images representing pathologies of the digestive tract.