A new method of brightness correction for multiview images

Significant brightness deviations between different camera views can often be observed in multi-view images. In this paper, a new method of brightness correction based on dynamic programming with control points constraint is proposed. The SIFT feature is firstly extracted and matched. The fundamental matrix constrain is then applied with RANSAC algorithm, and the inliners of matched points are acquired. Then, the brightness relationship between the matched points is extracted. After monotonic increasing consistence check, control points are finally acquired. In the process of dynamic programming based histogram matching, those control points are used as a constrain policy to improve the accuracy and reliability of matching results. Experimental results show that it can correct the brightness difference of multi-view images effectively, and can achieve better performance compared to Cox´s method.