Performance comparison between concatenated and ensemble learning approach.

<p>(a) shows classification error (%) of classifiers when trained with various texture features (LBP, LPQ, BSIF, GLCM, Histogram and Perception-like features), concatenated feature set and ensemble learning. These results are presented with both weak and strong cross validation. (b) and (d) show confusion matrix and its corresponding ROC curves for the concatenated features respectively. The mean AUC for CA and TA_LG is 0.908. (c) and (e) show confusion matrix and its corresponding ROC curves for the ensemble learning respectively. The mean AUC for CA and TA_LG is 0.960.