基于总变分彩色图像恢复问题的有效算法

An Effective Algorithm Based on Total Variational Color Image Restoration Problem

为改进彩色图像的恢复效果,针对数字图像在获取和传输过程中产生的图像退化问题,提出一种改进的总变分正则化模型.首先在最大后验估计的框架下,将彩色图像退化问题转化为总变分最小化问题;然后选择L1范数作为总变分模型的正则项;最后引入对偶变量,将上述问题转化为极大极小问题,利用一阶原对偶算法结合分块矩阵求逆的算法处理上述极大极小问题.实验结果表明,与交替迭代算法相比较,该算法对彩色图像进行去噪和去模糊的能力更优,实验验证了该算法的有效性和优越性.

In order to improve the recovery effect of color image,an improved total variational regularization modelis proposed to solve the problem of image degradation caused by digital image acquisition and transmission. Firstly,the problem of color image degradation is transformed into the total variational minimization problem under theframework of maximal posteriori estimation. Then the L1 norm is chosen as the regular term of the total variationalmodel. Finally the dual variables are introduced to transform the above issue into maximum and minimum that issolved utilizing first order primal dual algorithm,combined with inverse algorithm of block matrix. Experimentalresults show that compared with the iterative algorithm,the algorithm has better ability to denoise and deblur thecolor image. The effectiveness and advantage of the algorithm is validated by experiments.