【摘要】 传统的松弛方法有两个基本的更新策略：ｗｉｎｎｅｒ－ｔａｋｅ－ａｌ和ｌｏｓｅｒ－ｔａｋｅ－ｎｏｔｈｉｎｇ．这两个策略有以下缺陷：非一致性，局部极值问题和计算复杂性问题．作者将通过构造动力学系统来解决这些问题．这一思路的困难在于在逐步去掉不确定匹配的同时稳定系统．为此引入了特殊的非线性变换来构造这个动力学系统．与前人的工作相比，作者的算法较好地解决了计算复杂性和匹配准确率之间的矛盾．更多还原

【Abstract】 The existing updating strategies for the classical relaxation technique have various drawbacks such as nonuniformity, local minimum, and computational complexity. The authors design dynamic systems to deal with these problems. The main difficulty of this idea is how to simultaneously stabilize the system and obtained unambiguous match. Simple nonlinear transformations are introduced to construct the dynamic relaxation system for image matching. Compared with previous methods, The above mentioned approach effectively overcomes the contradiction between matching accuracy and computational complexity.更多还原