Image Registration is the vital image processing techniques to find precise pixel-to-pixel matching of two different images taken using the similar sensor at different time or dissimilar sensors at the same time/different time. By registering two images we can find the difference between the two images and also we can fuse information from the images. In this paper, automatic image registration of satellite images is performed by various feature extracting algorithms. The key points of each image are extracted using various feature extracting algorithms such as Speed Up Robust Feature (SURF), Minimum Eigenvalue, Harris, and then matching points are developed using Random Sample Consensus algorithm. For completely removing the outlier from the image and fitting the transformation model RANSAC has been used and the Neighbor-Distance-Ratio have been utilized as the feature matching- strategy. Finally, the image is registered using the affine transformation and the steps in image registration are fully automatic. The results are discussed based on the keypoints selection, matching points, and the number of outliers removed. The evaluation metric is based on computation time, matching ratio, Repeatability and RMSE.
Volume 12 | 07-Special Issue
Pages: 545-552
DOI: 10.5373/JARDCS/V12SP7/20202138