Active contour based segmentation of low-contrast medical images

The content of the paper is image processing with a special focus on image segmentation techniques. Frequently, there is need to work with images which are understandable only when some higher level knowledge is applied. Knowing these preconditions the search space can be reduced, which effects in time savings and better segmentation quality. The idea of the work is to start with the possible best initial approximation-the deterministic algorithms are employed, and to make this rough estimation converge to appropriate shape-it is attained by applying stochastically deformable contours. The tested images are the radiographic photographs (dental panoramas of teeth), which vary in terms of brightness and contrast. Therefore, the stage of image normalization and preprocessing is introduced.