Abstract: PET (Positron Emission Tomography) plays an increasingly important role in the research area of medical image processing. It抯 PET that first introduces quantitative analysis into the field of computer diagnosis. In the heart image of PET, contour fitting is one of the key steps in measuring the medicine absorptivity of the heart. In this paper, we concentrate on the problem of heart image segmentation. We present two different methods to solve this problem — mathematical morphology and spline model. They both get satisfying results in some aspects. The first method,which uses erosion and dilation the two basic operations to process the image, is simple but inflexible, while the second one is robust but time-consuming, which employs a set of spline curves to model the contour of heart, then transforms the original problem to a optimization problem. To seek the global optimal solution, we adopt the Simulated Annealing algorithm (SA), which is a stochastic relaxation technique. In the model, we also present an unique energy function. After solving some technical problems in the implementation of SA, we give comparisons between the two ways mentioned above.