This paper proposes a measurable criterion called "stitching reliability" which reflects the stitching quality of panoramic mosaics. We also show that the accurate focal length from image sequences closely relates to the stitching reliability. However, current methods to recover the focal length are not very robust because mis-stitched areas in general panoramas are used in the computation of the focal length. In our method, the several images with high stitching reliability can be automatically selected to speed up the focal length computation. To measure the stitching reliability, our algorithm computes the difference between real and ideal position of "image intersection point" for a camera rotational motion constraint. It is formally proved that the constraint is always satisfied in camera free rotation. Experiments show that the proposed method provides less deviation of focal length than others, and thus improves the quality of panoramic mosaics.
|Number of pages||4|
|Journal||Proceedings - International Conference on Pattern Recognition|
|Publication status||Published - 2000|
ASJC Scopus subject areas
- Computer Vision and Pattern Recognition