ABSTRACT
This chapter presents a novel procedure for automatic bone tissue segmentation using the spiral optimization (SO) strategy and a Gaussian modeled thresholding. As a pre-processing stage, a set of directional filters are applied over the computer tomography images in order to enhance and remove differently oriented features. In the segmentation stage, the SO strategy is used to initialize region-growing operators according to its intensity over the Gabor response image, and two Gaussian models are created from its result. Then, based on the intersection of these models, a thresholding strategy is introduced to classify bone and non-bone tissue pixels. Gabor filters have been widely used to characterize and extract texture features; they have also showed good performance for medical image processing applications, such as: microcalcification, edge, and tumor detection in mammographies, and segmentation of coronary arteries, just to mention some of them.
