A deformable template model based on fuzzy alignment algorithm

Abstract

A deformable template model for object extraction is proposed based on the fuzzy alignment algorithm (FAA). This object matching algorithm is partitioned into two iterative processes, the first is to estimate the pose relationship (point correspondence and transform parameters) between the current template and the prototype using FAA, the second is to adjust the current template under the exertion of internal energy and external energy functions. An affine-invariant internal energy function of the deformable template is utilized to deal with the transformation of the templates between different domains. Comparative studies with G-Snake model demonstrate the effectiveness of the proposed algorithm and show that it outperforms G-Snake in matching objects with large shearing of shapes.