Color–Texture Pattern Classification Using Global–Local Feature Extraction, an SVM Classifier, with Bagging Ensemble Post-Processing

This paper introduces a novel method for pattern classification, particularly relevant in image analysis applications requiring accurate classification of intricate patterns involving color and texture. The proposed method involves several key steps: the division of each image into global and local samples, extraction of texture and color features using Haralick statistics and a binary quaternion-moment-preserving method, classification using a support vector machine, and a final post-processing stage employing a bagging ensemble. A notable contribution is the image partition, which distinguishes global and local features, capturing a comprehensive representation of information for colored texture classification and yielding improved results. The method was rigorously tested on widely-used color–texture classification databases (Brodatz, VisTex, Outex, and KTH-TIPS2b), achieving correct classification rates of 97.63%, 97.13%, 90.78%, and 92.90%, respectively. Incorporating the post-processing stage further enhanced these results to 99.88%, 100%, 98.97%, and 95.75%. Comparative analysis with the best previously published results on the same databases demonstrated significant improvements in all cases, affirming the effectiveness and advancement of the proposed method.