Time-varying lighting conditions cause temporal variations of image intensities and
make most existing image matching techniques ineffective. As a solution to this irregular
lighting condition problem, we propose a novel matching technique based on gradient
orientation information in place of conventional image features such as intensities and
gradients. Gradient orientation is known to be insensitive to variations of intensities. Gradient
orientation information is utilized using two patterns that are obtained as the x and y
components of unit gradient vectors. Simulation results show that the proposed matching
technique is remarkably robust to both spatially uniform and non-uniform changes of image
intensities.