Stitching and Reconstruction of Linear-Pushbroom Panoramic Images for Planar Scenes

Abstract. This paper proposes a method to integrate multiple linearpushbroom panoramic images. The integration can be performed in real time. The technique is feasible on planar scene such as large-scale paintings or aerial/satellite images that are considered to be planar. The image integration consists of two steps: stitching and Euclidean reconstruction. For the image stitching, a minimum of five pairs of non-collinear image corresponding points are required in general cases. In some special configurations when there is column-to-column image correspondence between two panoramas, the number of image corresponding points required can be reduced to three. As for the Euclidean reconstruction, five pairs of non-collinear image corresponding points on the image boundaries are sufficient.

• Type: Conference
• Year: 2004
• Where: ECCV
• Authors: Chu-Song Chen, Yu-Ting Chen, Fay Huang
• Keywords: Computer Vision, ECCV 2004, Euclidean Reconstruction, Image, Image Corresponding Points