Image rectification is an essential part in most multi-view systems. It simplifies the stereo matching process by limiting the search path to horizontal and parallel lines. The main problem in conventional image rectification is complexity and mass computation load for calculation of calibration matrix, projection matrix and fundamental matrix. In this paper an efficient and confident method is presented to rectify stereo images based on the correspondence field (CF) that avoids the mentioned matrices. In general this algorithm is able to generate an image for any new camera position. Especially one can choose the new position in such a way that the virtual camera generates a rectified image. Both real and simulated stereo images are used to evaluate the proposed algorithm. SIFT algorithm is used to estimate how close is the generated image to original rectified one. Simulation results show the mean absolute error and standard deviation are about 0.25 and 0.26 pixels respectively and rectified image is independent of focal ratio.
Rights and permissions | |
![]() |
This Journal is an open access Journal Licensed under the Creative Commons Attribution-NonCommercial 4.0 International License. (CC BY NC 4.0) |