Title :
Distance Transform Approximation with a Multilayer Perceptron for 2D Image Recognition
Author :
Rodríguez-Ponce, Héctor-Uriel ; Rosas-Romero, Roberto
Author_Institution :
Departamento de Ingenieria Electronica, Univ. de las Americas, Puebla, Mexico
Abstract :
This paper presents a multilayer perceptron which approximates the distance transform of 2D images, specifically capital vowels. Approximation of images’ distance transforms are used with the purpose of fast recalling in well handwritten capital vowels within a square of 64x64 pixels. Architecture that has both, fast training and good approximation of distance transform is described. Possible improvements for further research are described too.
Keywords :
approximation theory; handwritten character recognition; image recognition; learning (artificial intelligence); multilayer perceptrons; transforms; 2D image recognition; distance transform approximation; fast training; handwritten capital vowels; multilayer perceptron; Artificial intelligence; Artificial neural networks; Equations; Humans; Image edge detection; Image recognition; Multilayer perceptrons; Neural networks; Pixel; Transforms;
Conference_Titel :
Electronics, Communications and Computers, 2005. CONIELECOMP 2005. Proceedings. 15th International Conference on
Print_ISBN :
0-7695-2283-1
DOI :
10.1109/CONIEL.2005.37
