In this section, we will discuss the segmentation techniques for gray-level image. This is because the subjects in our study, the MR images, are gray level intensity based, with pixel intensity within range 212-216 defined by different MR scanner manufactures. In addition, the methods for gray-level image are usually the basis for processing of a MR image sequence and multiple contrast images.
Gray-level image segmentation techniques have been studied for years. Among the existing algorithms in literature, some are based on the pixel intensity distribution or histogram [49-52], some use region-based splitting/merging approaches [11-14], and some are derived from morphological operations [53, 54]. They have been successfully employed in many applications. However, the drawback of these algorithms is the poor performance in noisy environment. Some Bayesian inference based segmentation techniques [19-22, 55], using the MRF as image model to improve robust performance to noise, have been proposed in recent years and become very popular.
This section will focus on the MRF model and its application on gray-level image segmentation. An enhanced version of the Highest Confidence First algorithm is introduced.
Was this article helpful?