Medical Image Segmentation Techniques, a Literature Review, and Some Novel Trends

Abstract

Segmentation requires the separation or division of an image into regions of similar properties. Image amplitude is the most basic attribute for image segmentation. Image texture and edges are also useful properties for the segmentation process. There is no standard approach for segmentation of an image; no single theory for image segmentation. Segmentation of an image is usually used to mark and determine boundaries and objects (curves, lines, etc.) in an image. More precisely, image segmentation is the process of labeling of every pixel in the image where pixels having the same properties have the same visual properties and share the same group. The result of segmentation process is a number of regions or segments that cover the whole image, or a number of extracted edges and contours of the image. All pixels in the same region are similar according to some characteristics or properties, such as texture, intensity, or color. In this paper a literature review of the various segmentation methods that are available for medical images is presented. Because of image segmentation importance, a set of image segmentation techniques namely; Thresholding techniques, Clustering techniques, Artificial Neural Networks, Edge based techniques, Region based techniques, Watershed, Graph based and Deformable models have been discussed and compared. The features and requirements of several freely and commercial software tools for image segmentation are clarified. The paper is ended by focusing on the novel trends on the topic.