Matching of complex patterns by energy minimization

Abstract

Two patterns are matched by putting one on top of the other and iteratively moving their individual parts until most of their corresponding parts are aligned. An energy function and a neighborhood of influence are defined for each iteration. Initially, a large neighborhood is used such that the movements result in global features being coarsely aligned. The neighborhood size is gradually reduced in successive iterations so that finer and finer details are aligned. Encouraging results have been obtained when applied to match complex Chinese characters. It has been observed that computation increases with the square of the number of moving parts which is quite favorable compared with other algorithms. The method was applied to the recognition of handwritten Chinese characters. After performing the iterative matching, a set of similarity measures are used to measure the similarity in topological features between the input and template characters. An overall recognition rate of 96.1% is achieved.

Segmentation and recognition of Chinese bank check amounts

“…The second data set was collected in order to train up and test the character recognizer [Leung 1997[Leung , 1998]. The set contains 3,735 Chinese character samples (83 samples for each of the 45 character classes).…”

“…The character recognizer used is a combined statistical and structural recognizer [Leung 1997[Leung , 1998]. The statistical recognizer extracts a set of 38 features from the unknown input and matches with each of the 45 character classes.…”

“…It is an iterative procedure and an energy function is used to guide the matching process. Details of the method are given in [Leung 1998]. The distance scores are entered into a table called the segmentation lattice and are used for subsequent parsing and final decision.…”

Matching of uncalibrated stereo images by elastic deformation

“…Matching is achieved by elastically deforming the two sets of segments. An energy function (Durbin and Willshaw, 1987;Leung and Suen, 1998) is defined such that, by minimizing this function, the segments of one pattern are iteratively moved toward the segments of the other pattern until most of the corresponding segments are aligned, as shown in Figure 3 as Patterns A and B. The energy function for guiding the movements of the segments of Pattern A toward the segments of Pattern B is given by …”

A Novel Fusing Algorithm for Retinal Fundus Images

“…The objective is to achieve local alignment while to maintain the regional structure as much as possible. We elaborately create an energy function whose original format can be found in [8] to guide the deformation process. For details of how to implement the local deform matching process, refer to [8].…”