Image Stitching using Harris Feature Detection and Random Sampling

In this paper the problem of fully automated panoramic image stitching for 2D image is introduced. In this work, stitching two images using invariant local features is used. This algorithm recognises multiple panoramas in an unordered image dataset and uses Harris Corner Detection for detecting the key point i.e. features. RANSAC method is used to choose the closest match between the two images by separating inliers and outliers. For the Image stitching using the inliers, homography matrix is used which requires least 8 feature points. Once the homography matrix is calculated between the two images, a panoramic view by wrapping the two images is generated. To get the efficient homography matrix, the feature points which fall onto the corresponding epipolar lines are selected. In this rectilinear projections are used to project the resulting image. In rectilinear projections images are viewed on two dimensional planes. Harris Corner Detection method is more robust for detecting the corners in the images than other methods. Hence, Harris Corner Detection with RANSAC which gives efficient image stitching