A systematic performance evaluation of interactive image segmentation methods based on simulated user interaction

In this paper we report on the results of a systematic performance evaluation of three efficient image segmentation algorithms. namely Graph-Cuts, Random-Walker and Grow-Cut. The evaluation focuses on their function as the computational part of an interactive segmentation system. The implications caused by the human involvement in the overall process are avoided by simulating two different patterns of user interaction. The methods are evaluated with respect to accuracy, precision, efficiency and parameter sensitivity on three dimensional medical images. The results provide useful insight regarding the algorithmic performance of the selected techniques and the effect of the identified patterns of user interaction on the segmentation outcome.