Multimodal image registration algorithm based on Shape Context and scale-invariant feature transform

An image registration method for multimodal images is proposed. This method is a combination of modified scale-invariant feature transform(SIFT) feature extraction algorithm and Shape Context feature descriptor. Feature points of multimodal images are extracted by modified SIFT feature extraction algorithm. Then each feature point is described by modified shape context descriptor, which forms a feature vector from the orientation histograms of sub-region around each feature point. After feature points matching with Euclidean distance as measure, the RANSAC algorithm is used to eliminate wrong corresponding pairs. At last, multimodal images registration is achieved by affine transformation and bilinear interpolation. Experimental results for registration of infrared images and vis