On measuring and employing texture directionality for image classification

Directionality is useful in many computer vision, pattern recognition, visualization, and multimedia applications since it is considered as an important pre-attentive attribute in human vision. To support using directionality (i.e., orientedness) for texture discrimination, a new measure that uses both local and global aspects of texture, with such use, to our knowledge, novel vis-à-vis prior state-of-the-art, to determine the directionality status for a texture is described and validated in this paper. This paper has four major elements. Element one is the measure we have developed that examines both local and global aspects of directionality to signal if a texture is directional or not. The local aspect is provided mostly from local pixel intensity differences, while a frequency domain analysis provides most of the global aspect. Element two is a comparison study of the measure (which exhibits the best outcomes) versus the known alternatives for determining texture directionality. Element three considers the measure relative to human experience. Element four considers applications of the measure to image classification. The second element (i.e., the study) is a comprehensive comparison study of existing texture directionality measures, based on the full set of Brodatz textures and human sentiment, which is the first such study.