BDBComp
Parceria:
SBC
Texture Classification Using Prediction by Partial Matching Models

Tatiane Cruz de S. HonórioLeonardo Vidal BatistaRodrigo C. Marques Duarte

This paper presents a new texture classificationmethod using histogram equalization and thestatistical modeling capability of the losslesscompressing algorithm Prediction by Partial Machine(PPM). The texture samples are preprocessed usinghistogram equalization. In the learning stage, thePPM algorithm constructs statistical models for thehorizontal and vertical structure of each class. In theclassification stage, texture samples to be classifiedare encoded by using the PPM models constructed inthe learning stage, in vertical and horizontal scanningorder. A sample is assigned to the class whose modelminimizes the average between the horizontal and thevertical coding rate. The classifier was evaluated forvarious sizes of samples and training set.

http://www.lbd.dcc.ufmg.br/colecoes/wvc/2009/0033.pdf

Caso o link acima esteja inválido, faça uma busca pelo texto completo na Web: Buscar na Web

Biblioteca Digital Brasileira de Computação - Contato: bdbcomp@lbd.dcc.ufmg.br
     Mantida por:
LBD