Comparison of Segmentation Methods in MR Image Processing

Image processing in biomedical applications is strongly developing issue. There were described many methods and approaches for image preprocessing, segmentation and visualization. It is necessary to choose suitable segmentation method to create a correct threedimensional model. The accuracy of reconstruction depends on precision of regions boundary determining in MR slices. The very often application is sensing of soft tissues. To imaging mentioned soft tissues is usually used tomography by magnetic resonance. Ideally, several tissue slices in three orthogonal planes (sagittal, coronal, transverse) are acquired. With slices in three planes is following reconstruction of shape of examined tissues most accurate. In case of acquired slices only in one plane the high spatial information lost occurs by image acquisition. Then it is necessary the shape of tissue appropriately reconstruct. At first the images are segmented and with use of particular segments the three dimensional model is composed. This article compares several segmentation approaches of MR images and their results. The results of segmentation by active contour, thresholding, edge analysis by Sobel mask, and watershed segmentation methods are compared. The results for different values of parameters of segmentation methods are compared. As the test image was chosen MR slice of human liver tumour.