Atsushi KOIKE

We address the problem of optimal color correction of multipoint camera systems that obtains dominant color pattern among all cameras. The color correction starts from any camera on the array sequentially until it reaches to the starting point and for several iterations. The iteration stops when the average of changes in color intensities is converged. The color correction is based on energy minimization using dynamic programming of a nonlinearly-weighted 2D Gaussian based kernel distribution of geometrically-matched correspondences, obtained by modified SIFT (Scale Invariant Feature Transform) method. The process is entirely automatic and experimental results show outperformance of the proposed method.