Paper Details

Image Registration is a fundamental image processing technique for aligning two or morepartially overlapping images of the same scene taken from different sensors, at different times, atdifferent depths, or from different viewpoints so that it can be made to form one panoramic imagecomprising the whole scene. The registration geometrically aligns two images (the reference and sensedimages). Due to the large image size with local distortion satellite image registration uses feature basedmethods. Scale-invariant feature transforms (SIFT) have been widely applied to register satellite images,which provide robust features. But the registration accuracy is affected by the lack of features and lowdistribution quality. This technique also suffers from a high-computational cost. A new feature basedimage registration technique is proposed in which the entire image X*Y be divided into smaller imageblocks M*N to process it in order to improve the accuracy as well as to extract more distinctive features.This paper proposes an automatic registration technique based on Scale Invariant Feature Transform(SIFT) features, which can deal with the large variations of scale, rotation and illumination of theimages.An outlier removal method called optimal RANSAC, which is an improvement of RANSAC isproposed; since standard RANSAC does not try to find the optimal set of inliers, instead it stops whenthe probability of finding more inliers reaches its predefined threshold. The consequence isthat it doesnot perform well for highly contaminated sets, i.e. when the set of outliers is large, since consensusmight not be reached within reasonable time.

AnkithaChandran A, Nelwin Raj N R, “FEATURE-BASED REMOTE-SENSING IMAGE REGISTRATION USING SIFT AND OPTIMAL RANSAC FOR OUTLIERS”, International Journal of Advance Engineering and Research Development (IJAERD), Vol. 1, Issue 7, pp. 124-129, July 2014.