Reliable image matching based on relative gradients

We present an image alignment algorithm based on the matching of relative gradient maps between images. This algorithm consists of two stages; namely, a learning-based approximate pattern search and an iterative energy-minimization procedure for matching relative image gradient. The first stage finds some candidate poses of the pattern in the image through a fast search of the best match of the relative gradient features from the database of training feature vectors. The training database is obtained from the synthesis of the template image under a number of uniform samplings in a range of the geometric transformation space. Subsequently, the approximate candidate poses are further verified and refined by matching the relative gradient images through an iterative energy-minimization procedure. This approach based on the matching of relative gradients has the advantage of robustness against inhomogeneous illumination variations. Some experimental results are shown to demonstrate the efficiency and robustness of the proposed algorithm.