Research And Application Of Underwater Image Enhancement Technology

With the increasing applications in the underwater imagery field,underwater image enhancement techniques are becoming increasingly important.Many underwater studies and applications now utilize digital imaging and computer vision techniques,and they are vital to obtain clearer images,where image enhancement is a very important part of this process.Underwater images often suffer from color cast and low contrast due to the refraction and scattering of underwater light,especially for challenging underwater images,which are often followed by a color restoration operation after dehazing.It is difficult to judge the final enhancement effect only according to human subjective vision,and some underwater image evaluation criteria are often used to objectively assess the image enhancement effect.In order to evaluate the image enhancement effect more objectively,this thesis first briefly introduces the classical underwater image evaluation criteria: UIQM and UCIQE,and then analyses their advantages and disadvantages.This thesis additionally introduces a recent evaluation criterion for the light source estimation,i.e.,the reproduction angle error.In addition,this thesis optimizes the reproduction angle error to improve its computational efficiency.The blurry characteristics of underwater images are similar to those of foggy images on land,and the patch-based dark channel prior(DCP)method is often used when enhancing underwater images.In order to remove the patch artifacts in these DCP-based enhancement methods,a coordinated underwater dark channel prior(CUDCP)method is proposed in this thesis.The proposed method uses adaptive transmission map estimation,and uses the compensated gray world theory for color restoration and finally performs the differentiated patch-division average filters based on sub-patch classification.The experimental results show that compared to the state-ofthe-art enhancement methods,the proposed method can significantly reduce patch artifacts in challenging image enhancement,while maintaining the objective quality of these underwater images and improving their subjective quality at a reasonable computational cost.In addition,by combining the above work and related technologies,this thesis develops an underwater image enhancement system.The system brings together multiple algorithms for single image enhancement,image set enhancement,image evaluation criteria,video enhancement,error concealment and other functions.It provides certain facilities and auxiliary tools for people to study the underwater imagery.