Texture Image Categorization in Wavelet Domain via Naive Bayes Classifier Based on Laplace and Generalized Gaussian Distribution

In this paper, we have investigated recently proposed feature extraction technique for texture image representation. In the introduced method, features are extracted via bounded Laplace mixture model (BLMM) in wavelet domain. Due to nature of wavelet coefficients that can be modeled accurately with Laplace distribution, it is proposed to apply classifiers based on this distribution, which leads us to introduce Naive Bayes classifier with Laplace distribution for image categorization. The proposed approach is validated through experiments on different texture image datasets and it has shown very good results as compared to the model based on Gaussian distribution. The generalized Gaussian distribution is a generalization of both Laplace and Gaussian distributions, thus we have introduced also Naive Bayes classifier with generalized Gaussian distribution to achieve better performance as compared to the above two models. The proposed approach is also validated through extensive experiments and it is observed that by taking into account the nature of data, proposed models have very good performance. Classification results are presented by different performance metrics to ensure the effectiveness of proposed algorithms in texture image classification.