一种细节保持的Retinex图像增强算法

Novel detail preserving Retinex algorithm for image enhancement

针对基于Retinex理论的Kimmel变分模型在图像增强中存在人工伪影等问题,对其模型进行了改进.改进的模型增加了保持细节信息的因子,并在模型求解过程中采用了基于金字塔的提升算法,计算复杂度明显降低,然后对求解得到的图像进行正态截取拉伸处理,取得了较好的图像增强效果.

In image enhancement, due to the blurring problem of the Kimmel variation model based on the Retinex theory, a modification for the Kimmel model is presented. The new model modifies the Kimmel model by increasing an expression for preserving the detail, and then calculates the optimal illumination image by combining pyramid data structure and lifting algorithm, which reduces the computational complexity. Finally, the normal clipping stretch is deployed to process the reflectance image. Experiments show that this algorithm can effectively remove obscuration effect from color images.