雾霾图像清晰化算法综述

A review of hazy image sharpening methods

雾霾图像不仅影响视觉效果,而且模糊不清晰的图像容易为后续识别、理解等高层次任务带来困难。雾霾图像清晰化问题是一个典型的不适定问题,其成像过程难以精确建模,消除图像中的雾霾面临巨大的挑战。近年来,研究者提出大量的图像去雾算法克服雾霾引起的图像降质退化,为全面认识和理解图像清晰化算法,论文对其进行梳理和综述。首先,对雾霾图像清晰化算法进行整理,根据雾霾退化过程是否有模型支持,将清晰化算法分为基于Retinex模型、大气散射模型去雾算法和无模型图像去雾算法。大气散射模型是有模型算法中主流模型,本文详细剖析了模型成像机理,并根据其成像机制揭示大气散射模型容易受大气浓度均匀分布假设的限制,较难处理非均匀雾霾图像问题。基于深度学习的无模型图像去雾算法则不仅可以应对非均匀雾霾图像,而且去雾性能获得了极大地提升。其次,本文汇总了近年来常用去雾数据集,从数据集适应范围、规模、可扩展性等多个维度进行总结。并根据雾霾图像形成方式,对人工合成雾霾数据集和真实拍摄数据集分别从定性和定量的角度探讨了数据集对图像去雾算法的影响。

Hazy images not only affect the visual effect,but also easily bring difficulties to the subsequent high-level tasks such as image recognition and understanding.Image dehazing is a typical ill-posed problem and it is difficult to accurately model the imaging process,so eliminating the haze in the image faces enormous challenges Researchers have proposed plenty of methods to overcome the hazy image degradation caused by haze.This paper summarizes the image dehazing methods to fully understand and organize them.In general,image dehazing methods can be divided into model-based image dehazing and model-free image dehazing.Image dehazing based on the atmospheric scattering model is the mainstream model of model-based dehazing.We analyze the imaging mechanism of the model detailed,and reveal that the atmospheric scattering model is good at describing the hazy image when the atmosphere is homogenous.The dehazing methods based on deep learning can avoid the restrictions of atmospheric scattering model,they can effectively process all kinds of even and uneven haze images.Secondly,this paper summarizes the commonly used image dehazing data sets in recent years,and compares the data sets from multiple dimensions such as application,scale and expansibility.Moreover we discuss the effects of the data sets with different image dehazing methods from qualitative and quantitative perspectives respectively.