Abstract
In this paper, a multi-frame based homography estimation method is proposed for video stitching in static camera environments. A homography that is robust against spatio-temporally induced noise can be estimated by intervals, using feature points extracted during a predetermined time interval. The feature point with the largest blob response in each quantized location bin, a representative feature point, is used for matching a pair of video sequences. After matching representative feature points from each camera, the homography for the interval is estimated by random sample consensus (RANSAC) on the matched representative feature points, with their chances of being sampled proportional to their numbers of occurrences in the interval. The performance of the proposed method is compared with that of the per-frame method by investigating alignment distortion and stitching scores for daytime and noisy video sequence pairs. It is shown that alignment distortion in overlapping regions is reduced and the stitching score is improved by the proposed method. The proposed method can be used for panoramic video stitching with static video cameras and for panoramic image stitching with less alignment distortion.
Highlights
With the development of information and communications technology (ICT), there is a growing demand in various fields, such as virtual reality (VR), image security, entertainment, aerial image, and mobile applications for panoramic images and videos with a wide angle of view [1,2,3,4]
For a quantized location bin Gvr, if there exists more than one feature point in the location bin, a feature point with the largest blob response is selected as the representative feature point for the location bin in the homography estimation interval, as follows: pv =
The proposed method can improve the accuracy of homography by matching representative feature points of the largest blob responses and by replicating these matches for random sample consensus (RANSAC) according to their histogram counts and strengths of blob responses
Summary
With the development of information and communications technology (ICT), there is a growing demand in various fields, such as virtual reality (VR), image security, entertainment, aerial image, and mobile applications for panoramic images and videos with a wide angle of view [1,2,3,4]. The centroids [22,23,24] or motion trajectories [25] of moving objects in different cameras are used as feature points for homography estimation These methods require solving object tracking and data association problems across camera and cannot be applied to video sequence pairs without moving objects. The corresponding pixels in a different view that has the most similar activity features are selected for matching These methods are robust against arbitrary orientations and zoom levels of cameras and illumination conditions, they have the limitation that moving objects must exist in the overlapped regions in the video sequence pair. To improve the quality of stitched video sequences, it is necessary to find correct alignment parameters between cameras while suppressing the wrong feature point extraction from time-varying noise.
Sign-in/Register to view highlights & summary Sign-in/Register to access full text options 
Free) 
Talk to us
Join us for a 30 min session where you can share your feedback and ask us any queries you have
Schedule a call
