Two-phase Relaxation with Fuzzy Split-and-merge Contraption

• Main Authors: Chen, Jin-Jinn, 陳金進
• Other Authors: Chen, Sei-Wang
• Format: Others
• Language: zh-TW
• Published: 1995
• Online Access: http://ndltd.ncl.edu.tw/handle/28474231878335434177

碩士 === 國立臺灣師範大學 === 資訊教育研究所 === 83 === 　　We endow general structural relaxation schemes with a number of novel features: (a) the capability of invariance under affine transformations (including rotation, translation, and scaling), (b) the competence of handling multiple objects with incomplete shapes, (c) a two-way framework in which both consistency and correctness of feature mapping are improved, and (d) a fuzzy split-and-merge mechanism that significantly reduces the computational complexity. Experimental results have revealed the robustness and efficiency of the proposed scheme in dealing with multiple-object scences. The outcomes of the algorithm are good enough for being used in the task of high level fearture matching in object recognition. 　　Index Terms-Structural relaxation scheme, Affine transformation, Mutual occlusion, Two-way relaxation, Fuzzy split-and-merge contraption, Subtended angles