360-Degree image stitching with GPU support
Fanni Molnar, Andras Kovacs · 2016
Nowadays, in the field of observing large indoor and outdoor areas there is growing use of automated electronic solutions. These observations are usually carried out by one or more multi-camera systems. In this paper, our goal is to construct a very similar system but with the usage of only one camera. The system we propose will be able to analyze the panoramic pictures made from the visual data given by a rotating camera. The main subject of this paper is the stitching of images - in a simple, rapid and efficient way - made by a rotating camera and after optimizing the process in a data-parallel environment. This algorithm we propose differs from the popular, feature-based stitching pipelines. There is no keypoint detection and description like SIFT or SURF or ORB. We approach from a different angle to accomplish the mentioned goals.