Comparison and Application of Implementing Image Homographs in Computer Vision
Xingqi Qiu · Advances in computer science research · 2024
In the field of computer vision, planar homography plays a pivotal role in our research process.The homography matrix is capable of performing a variety of functions such as image warping, stitching, and video stitching.Within the realm of epipolar-geometry, it enables the execution of numerous tasks, including 3D reconstruction.This paper primarily focuses on the creation of panoramic images through automatic stitching of photographs with using homographic matrix, comparing the efficacy and efficiency of different feature extraction algorithms in terms of feature point matching, like Speeded Up Robust Features (SURF), Features from Accelerated Segment Test (FAST), and KAZE Scale-Invariant Feature Transform (SIFT), Oriented FAST and Rotated BRIEF (ORB),and put forward some applications.Consequently, this leads to variations in the effectiveness and efficiency of images stitched using the homography matrix.This paper finished a feature matching experiment based on comparing the panoramic image with using different feature detect algorithms.For scenarios requiring high accuracy where processing time can be longer, SIFT, KAZE, or SURF might be better choices.On the other hand, for applications that need fast response, FAST or ORB would be more appropriate.