Paper Info
Title
A New Thinning Algorithm Based on Controlled Deletion of Edge Regions
Abstract
In this paper a new thinning algorithm for binary patterns is presented. The algorithm is based on an iterative controlled removal procedure working on entire regions of the pattern. The thinning process permits to obtain skeletons retaining important specificities of the pattern useful for robust structural descriptions. The experimental results carried out on typewritten and handwritten characters point out the efficacy of the algorithm with respect to other techniques in literature. The thinning of a binary pattern is a widespread process in data transmission, image processing, and pattern recognition. It consists in the successive removal of contour pixels until the pattern is reduced to a line drawing.1 In the past precise definitions of digital skeletons have been proposed based on complex mathematical frameworks2,3 and, in order to meet the needs of different applications, many different thinning algorithms based on both sequential and par­ allel approaches have also been developed.4-13 In the field of data transmission, for example, the main advantage of thinning is to reduce the amount of data required for transmitting the pattern without loss of shape information, while in the field of structural pattern recognition it is to retain the characteristics of the shape useful for the specific structural description.14,15 In this paper a new thinning algorithm producing four-connected skeletons is presented. It works sequentially on the contour pixels of a binary pattern but it performs deletions of entire regions of pixels simultaneously using specific local information. The skeletons exhibit a very robust representation of local specificities of the shape and are not very sensitive to pattern distortions. In this sense the algorithm is effective for character recognition systems based on structural tech­ niques in which structural elements of the pattern are identified taking into account specific points of the skeleton as bend-points and cross-points. The test of the algorithm has been carried out on both typewritten and hand­ written characters and the skeletons have been compared with those produced by the Safe Point Thinning Algorithm (SPTA).4 The results point out the efficacy of the new algorithm due to its small sensitivity to handwriting variability and to the adaptability of the deletion process to different shapes.
Year
DOI
Venue
1993
10.1142/S0218001493000492
IJPRAI
Keywords
Field
DocType
sequential thinning process,binary pattern,structural analysis.,four-connected skeleton,structural analysis,data transmission,pattern recognition,structure analysis
Binary pattern,Pattern recognition,Thinning,Removal procedure,Artificial intelligence,Mathematics,Thinning algorithm,Binary number
Journal
Volume
Issue
Citations 
7
5
5
PageRank 
References 
Authors
0.53
14
3
Name
Order
Citations
PageRank
Giovanni Dimauro119024.09
Sebastiano Impedovo221726.47
Giuseppe Pirlo323434.30