The invention discloses be based on (RANSAC) method of sampling improvement of CSP random sample consensus with homography precision in augmented reality. Sampling method estimate homography display conversion multiple images between relationship, using RANSAC algorithms include in the pre-treatment step of image which image be divided into multiple dot matrix regions, and therefrom extract image characteristic point be divided into lattice region; It is applied to selection facial feature estimation homography with the characteristic point of above-mentioned selection second step, all disclosed constraint classification. ;The 2014 of copyright KIPO submissions;[Reference numerals] (AA) (one): the characteristic point initially selected, (second): the limitation of distance violating the regulations, (3): violating linear limitation, () () () (d e f g): selection characteristic point
展开▼