基于先验知识与大气散射模型的图像增强算法

Image Enhancement Based on Prior Knowledge and Atmospheric Scattering Model

针对现有图像增强算法大多不具备处理多种类型降质图像的能力,提出一种基于先验知识与大气散射模型的快速图像增强算法.首先,通过大量实验统计,提出一种新的图像先验—明亮通道先验,即高质量清晰图像中每个像素邻域都极有可能存在白点;随后,对散射模型所存在的缺陷加以改进,并结合明亮通道先验与黑色通道先验,推导出场景反射率的恢复公式;最后,针对黑色通道先验失效情况,提出一种基于可靠性预测的容错机制,以提高其适用范围.实验结果表明:本文算法不但可以有效的突出纹理细节,还具有一定的色调恢复功能,能够处理多种不同类型的降质图像.

Almost the existing image enhancement algorithms are not capable to deal with multi-type degenerated images. In this paper, a fast image enhancement algorithm is proposed based on prior knowledge and atmospheric scattering model. Firstly, an innovative image prior, which were called bright channel prior, were proposed through experimental statistics. It shows that the whites spots are more likely to exist in the neighborhood of each pixels in high-quality clear image. Then, the scattering model were improved and it derived the calculation formula of reflectance image by bright channel prior, dark channel prior and improved scattering model. Finally, we put forward a fault tolerance mechanism, which is based on reliability prediction, to handle the situation that the dark channel prior does not work. The experiment results shows that this algorithm can not only highlight the texture details well but also restore the tone. It is able to deal with multi-type degenerated images.