Fusion of Gabor and co-occurrence texture features for document image regions classification

In this work we examine two categories of texture features and several classifiers for document image regions classification. We propose a reliablefeature vector using both Gabor and Co-occurrence texture features which is appropriate for Bayesian classifier. The proposed feature vector is evaluatedover a data base which contains 1000 text regions and 1200 non-text regions. Using Bayesian classifier we obtain true positive=98% and false positive=1.8% for the test set.

کلیدواژه ها:

region classification ، Gabor wavelet ، Cooccurrence matrix ، Bayesian classifier ، document segmentation