tailieunhanh - Báo cáo hóa học: " Region Information-Based ROI Extraction by Multi-Initial Fast Marching Algorithm"

Tuyển tập báo cáo các nghiên cứu khoa học quốc tế ngành hóa học dành cho các bạn yêu hóa học tham khảo đề tài: Region Information-Based ROI Extraction by Multi-Initial Fast Marching Algorithm | EURASIP Journal on Applied Signal Processing 2004 11 1739-1749 2004 Hindawi Publishing Corporation Region Information-Based ROI Extraction by Multi-Initial Fast Marching Algorithm Zhang Hongmei School of Life Science and Technology Xi an Jiaotong University Xi an 710049 China Email claramei@ Bian Zhengzhong School of Life Science and Technology Xi an Jiaotong University Xi an 710049 China Email bzzbme@ Guo Youmin First Affiliated Hospital Xi an Jiaotong University Xi an 710049 China Email Ye Min Institute of Mechanical Engineering Xi an Jiaotong University Xi an 710064 China Email minye2000@ Miao Yalin School of Life Science and Technology Xi an Jiaotong University Xi an 710049 China Email myl@ Received 23 March 2003 Revised 10 January 2004 Recommended for Publication by Kyoung Mu Lee Region of interest ROI plays an important role in medical image analysis. In this paper a new approach to ROI extraction based on the curve evolution is proposed. Different from the existent method the proposed approach is efficient both in segmentation results and computational cost. The deforming curve is modeled as a monotonically marching front under a positive speed field where a region speed function is derived by minimizing the new defined ROI energy and integrated with the edge-based speed function. The curve evolution model integrating the ROI information has a large propagation range and could even drive the front in low-contrast and narrow thin areas. Moreover a multi-initial fast marching algorithm which permits the user to plant several seed curves as the initial front and evolves them simultaneously is developed to fast implement the numerical solution. Selective planting seed curves could help the local growth and thus may further improve the segmentation results and reduce the computational cost. Experiments by our approach are presented and compared with that of the other methods which show .