Extraction and Tracking of Orientation Coded Features Being Robust against Illumination Changes

We present a feature extraction and tracking method which is based on pure rich texture analysis and has robustness under irregular conditions, such as illumination change, deformation of objects and so on. The method is composed mainly of two algorithms: Entropy Filtering and Orientation Code Matching (OCM). Entropy filter points up areas of images being messy distribution of orientation codes. The orientation code is determined by detecting the orientation of maximum intensity change around neighboring eight pixels. The areas being extracted by the entropy filter doesn’t depend on edges and gradations of images and has robustness against illumination change and motion of objects. Therefore, we can say the areas have ”pure rich texture” and are suitable for tracking based on visual information. And then, OCM, a template matching method using the orientation code, is applied to track templates being centered on the features and updates the templates each frame. OCM is speeded up by using the information of texture richness and motion of objects. Because of their large robustness, we can track the templates robustly under irregular conditions.