The automatic construction of large, high-resolution multi view image registration is an active area of research in the fields of image processing. Multiview image registration can be used for many different applications. The most traditional application is the construction of large aerial and satellite photographs from collections of images, construction of virtual travel etc. This proposed Automatic feature based image registration method does not allow any user interaction and perform all registration steps automatically. Here the matching points are found automatically using local feature detector i.e. harris corner detector which find invariant features using feature descriptors as oriented patches. For estimating homography between detected features of images to be registered, Homography estimator i.e. modified RANSAC (RANdom SAmple Consensus) algorithm, and direct linear transformation algorithm is used. Here features are located at Harris corners (new improved) in discrete scale-space and oriented using a blurred local gradient. To have better spatial distribution of features, adaptive non- maximal suppression algorithm is used.Feature matching are achieved using RANSAC which also uses DLT (Direct Linear Transformation) and warping is applied to achieve final registered image. This proposed algorithm can be applied for the series of images that may or may not be in the same alignment as per desired output image, thus mainly scaling, rotation and image transformation must be applied to get proper registered image.
Automatic Image Registration
