A NOVELTY APPROACH OF SPATIAL CO-OCCURRENCE AND DISCRETE SHEARLET TRANSFORM BASED TEXTURE CLASSIFICATION USING LPBOOSTING CLASSIFIER

Vivek Vivek

doi:10.3844/jcssp.2014.783.793

Abstract

Recently, the research towards Brodatz database for texture classification done at considerable amount of study has been published, the effective classification are vulnerable towards for training and test sets. This study presents the novel texture classification method based on feature descriptor, called spatial co-occurrence with discrete shearlet transformation through the LPboosting classification. It can be considered as a frame through the texton template that mapped into the texture images and it works directly on relating the adjacent spatial with its pixel boundary through the local intensity order. Hence, the proposed method for the feature extraction and classification of texture suggested with the experimentation through the spatial co-occurrence matrix with the power spectrum based discrete shearlet transform and it classified through the LP boosting method on Brodatz database images. This hybrid second order statistical based classification method significantly outperforms the existing texture descriptors the multiscale geometric tool shows the proposed method outperforms other classification method.

Highlights

Spatial gray tone co-occurrence probabilities with its spatial frequency in a texture based image are suggested
This study presents the novel texture classification method based on feature descriptor, called spatial cooccurrence with discrete shearlet transformation through the LPboosting classification
We propose the combination of multi-texton histogram with the Discrete Shearlet Transform (DST) to discriminate the Brodatz album based on feature extraction and it undergoes classification with the minimax theory based LPboost classifier the accuracy of this system very well compared to other state of art techniques

Summary

Introduction

Spatial gray tone co-occurrence probabilities with its spatial frequency in a texture based image are suggested. In our previous work focus primarily on the integrated the characterization of textures based on Discrete Shearlet Transform (DST) by extracting entropy measure and to classify the given Brodatz database texture image using K-Nearest Neighbor (KNN) classifier by Vivek and Audithan (2014) Such adaptation improves the classification accuracy, it severely increases the feature space complexity. We propose the combination of multi-texton histogram with the Discrete Shearlet Transform (DST) to discriminate the Brodatz album based on feature extraction and it undergoes classification with the minimax theory based LPboost classifier the accuracy of this system very well compared to other state of art techniques In this proposed approach, the Brodatz database image feature are analyzed through the spatial co-occurrence of local intensity order and the extracted features are decomposed into various patterns by discrete shearlet transform and the obtained pattern are classified based on the similarity through the LP boosting method.

Results

Discussion

Conclusion