Image Stitching Algorithm Based on Improved Harris Corners Extraction

This paper proposes an improved Harris corners extraction algorithm, which consists of coarse screening and fine screening in the corners extraction process. Firstly, image segmentation is used to find the redundant regions of feature points in the image. Secondly, the original image is divided into 3x3 blocks to separately deal with the response values of the Hession matrix of the corners contained in each block. Finally, for the corners in the three small blocks on the left of the figure, since they are the farthest from the overlapping area, only two corners with the largest response values are taken from each small block. Similarly, the corners in the middle three small blocks are required to take 20%-60% of the number of corners in each block, and the corners on the right three small blocks are required to take 90% of the number of corners in each block. After removing the redundant corners, the rest corners are the key feature points to be matched. Experimental results show that the algorithm proposed by this paper not only improves the operating speed, but also improves the registration accuracy by reducing the number of corners, so as to lay the foundation for subsequent image stitching.