This paper presents a quadratic programming based cluster correspondence projection (QPCCP) algorithm, where the optimal correspondences are searched via gradient descent and the constraints on the correspondence are satis-fied by projection onto