EVL research assistant and Electrical and Computer Engineering PhD student, Haoyu Wang presents his research contribution to the National Science Foundation sponsored “Sensor Environment Imaging (SENSEI) Instrument Project”.
The goal of SENSEI project is to create a scientific camera and display system for fully surrounding stereo cinema for scientific visual and depth data acquisition. Thus, the construction of 360 X 180 panorama are quite important for creating VR-like experiences based on the captured omnidirectional visual data. Right now, several different panorama generated methods have been investigated by the SENSEI research team:
point cloud reprojection
image stitching method
EPI technology and my task here is to implement an automatically panorama generation system with scripted-based panotools, which is belong to the image stitching method.
In this talk, the advantages and disadvantages of the point cloud reprojection and image stitching methods are compared. In addition, an improved depth-based matching, which utilizes the depth information to help qualify 2D corresponding features will be presented. In order to solve the vertical disparity problem between left and right panoramas, one stereo optimization idea is proposed and will explain how to eliminate the vertical disparity and guarantee audiences can perceive nice depth information from the stereo panorama. Finally, two stitched 360 X 180 panoramas from 72 images in spherical coordinate will be displayed.