A robust measure for visual correspondence

In this paper a novel measure for visual correspondence is proposed, to be adopted for common computer vision tasks such as pattern matching, stereo vision, change detection. The proposed measure implicitly exploits the concept of order-preservation between neighbouring pixels and it is suitable for those cases where disturbance factors such as photometric distortions and occlusions occur between the images to be matched. Furthermore, this measure tends to be robust in presence of significant amount of noise which can be introduced, e.g., by cheap camera sensors. Experimental results demonstrate the effectiveness of the proposed approach in a typical template matching scenario as well as in a particular application dealing with secure gate access control.