一种空间自适应的Retinex变分校正模型

针对传统Retinex变分模型采用相同的权重容易引起虚假痕迹的缺陷，通过引入差分特征值作为边缘指示算子，构造了一种具有空间自适应调节能力的Retinex变分校正模型，该模型能够利用影像空间域信息来控制变分校正模型在不同像素点的约束强度，在边缘区域施加较小的正则化约束保持影像的边缘特征；而在平坦区域施加较大的正则化约束。同时根据反射分量的物理性质，在变分校正模型中对其施加均值逼近灰度中值约束防止局部曝光过度。本文用分裂Bregman迭代法实现对该变分校正模型的最优化求解，利用模拟影像和真实影像进行实验，并与传统方法进行比较，结果表明，本文方法能够消除影像灰度不均匀现象，同时大幅提高计算效率。

A Spatially Adaptive Retinex Variational Model for the Uneven Intensity Correction

In order to overcome the disadvantage of the tradition methods that is easy to cause false traces with the same weight, a spatially adaptive retinex variational model by using the edge indicator called difference eigenvalue is proposed, which can automatically balance the regularized strength between different spatial property regions in the image. In the edges, weak regularized strength is enforced to preserve detail, and in the homogeneous areas, strong regularized strength is enforced to eliminate the uneven intensity. The gray world assumption is employed to constrain the reflectance, and a fast computation method is used to solve the proposed model based on the split Bregman iteration. Experimental results with both simulated and real-life data are given to show the effectiveness of the proposed model in terms of the quality of enhanced image and the efficiency of the minimization algorithm.