Computational optical distortion correction based on local polynomial by inverse model

An inverse model and computational optical distortion method is proposed. The method is applied to correct complex optical distortion in off-axis optical system, such as optical system found in head-mounted display. The presented method divides the field of view into several subsections. And each subsection is corrected separately using corresponding distortion parameters. Direct mapping between standard image displayed by head-mounted display and distorted image captured by camera is used to achieve distortion parameters instead of using discrepancy relationship between them. Mapping direction in this paper is from distorted image to standard image, which is different from conventional model which has opposite direction. Pixel filling is employed to calculate gray level. Inverse spatial mapping and pixel filling form inverse model. To avoid seam-line across adjacent areas concerning field of view division, a corresponding method called overlapping image method is used to eliminate discontinuity between mosaicking boundaries. Experiments including comparison between forward model and inverse model and different orders of polynomial were carried out on optical see-through head-mounted display to verify the effectiveness of our algorithm and the experimental results with high precision indicate the proposed algorithm is suitable for complex and asymmetric distortion correction of wide-angle, off-axis optical system.