Outdoor Scene Classification Using Local Binary Pattern and Its Variants: A Survey

In computer vision, image representation is a challenging issue in the field of texture recognition, which is an evolving area of research. This paper presents a comprehensive review of historical and recent state-of-the-art methods in the feature extraction field. Detecting the most appropriate texture descriptor allows us to perform various classification and predictive modeling problems. In this survey, we provide a taxonomy of different extensions as an effective preprocessing technique, studying their robustness and limitations. Moreover, the classification step is executed using five powerful supervised machine learning methods. Subsequently, experimental evaluation is verified on a large benchmark dataset for outdoor scene textures. Nevertheless, the Multiscale Local Binary Pattern learned by the Random Forest classifier attains a high level of performance in all terms. These findings and knowledge contribute to a faster, more cost-effective, and more comprehensive approach to performing texture analysis.