| 摘要 |
The method involves determining in a common zone (20) pairs of significant pixels in correspondence on each of two input images (P1, P2). A matrix type transformation (M1) between the two input images is determined, and a focal length is determined in view of rejoining. Perspective error correction is effectuated taking into account a depth component for each of pairs and a new matrix type transformation (M2) is determined. Cylindrical type rejoining is effectuated after collecting a set of pixels representing a superposition of input images. An independent claim is also included for an automatic panoramic reconstruction system. |
