梯度角约束图像插值
An Image Interpolation Method with Gradient Angle Constraints
-
摘要: 为了使插值的图像既保持光滑的图像轮廓又获得清晰的图像边缘,提出一种梯度角约束图像插值方法.首先证明了局部恒定的梯度角约束插值等价于总变分最小化;在此基础上,提出一个具有更强连续性的梯度角约束,由这个约束条件导出的三阶偏微分方程使插值的图像既可以保证水平线方向的连续性,又减小图像边缘的宽度;最后证明了插值方程中正则项的对比不变性.通过数值实验证明了该方法的有效性.Abstract: An image interpolation method with gradient angle constraints is proposed to obtain smooth image contours and crisp edges.First, awe have proved that the constraint of local constant gradient angles is equivalent to the minimization of total variation.Based on this, aa stronger continuity constraint of gradient angle is presented.The interpolation formula (a3-degree-PDE) derived from the constraint can ensure the continuity of level lines and reduce the width of edges.The contrast invariant of the regularizer of the formula is also proved.Numerical experiments demonstrate the validity of the proposed algorithm.
下载: