毕业论文移动版

摘要当空气中会悬浮着固态小颗粒或者是液态小水滴的浓度高时，景物中的光经过这些物体的折射和反射，形成雾霾图像。对于图像的处理造成很大的影响。

 本文首先介绍了前人对于图像去雾研究的一些算法，然后具体就暗通道先验算法进行了介绍，就透射率图和复原图像进行了优化，并以此作为课题研究的基础，

在研究了图像去雾的基础上，本论文探讨了视频的去雾算法。为了减少视频帧计算的复杂度，将关键帧透射率传递到非关键帧的透射率。为了进行帧间传递，首先研究了基于帧位置关系的线性插值，但是效果不太理想。然后在雾霾图像形成原理的基础上，又推导了帧图像之间的透射率的关系，直接求得的透射率细节处理的不好，并用导向滤波进行优化。但是这种方法没有考虑到视频图像的时空一致性信息，我们基于LK光流法，运用运动估计的方法，从原图像中求得运动矢量，基于运动矢量对于透射率图进行双线性插值，得到新的透射率图，为了解决遮挡问题，首先计算两个关键帧的透射率图，然后运用上述算法得到非关键帧的透射率图，运用像素匹配的方法，选择使用哪个透射率图，得到比较好的结果。82522

毕业论文关键字：图像去雾 光流法运动估计视频去雾

毕业设计说明书外文摘要

Title The Research of Video Dehazing Abstract When solid particles and liquid particles that suspend in the air have low

concentration, the air light will reflect and refract。This is the formation of fog and haze image。 And this will affect the image processing。

This paper introduces the defog method that based on image enhancement and image restoration firstly。We compare their advantages and disadvantages。Then we introduce one of the algorithms to better fog effect----dark channel prior algorithm。And we regard it as the basis of video dehazing。And we optimize the transmission map by using the guided filter。 We enhance the dehazed image by using histogram equalization to accord with visual effect。

We research the algorithm of video dehazing on the base of image dehazing。 We will transmit the transmission of key frame to that of non key frameto reduce the time complexity。Firstly,the linear interpolation based on the position relation of the frame is studied。But the result is not very well。And then we deduct the relation of frame based on the principle of fog and haze image formation。But the result is not very well,the details of map are very bad。Then we optimize the transmission map by using the guided filter。However these methods do not consider the time-space consistency of video image。We get the motion vector by using the LK optical flow,then we will get the new transmission by using thebilinear interpolation。In order to solve the occlusion problem,we calculate the key frame transmission map。Then we get the non key frame transmission by using that methods。And we select the transmission map by using pixel matching method。Then we can dehaze the video frame and we can get good result。

Keywords : image dehazingoptical flowmotion estimationvideo dehazing

1引言 1

 1。1研究意义 1

 1。2研究现状 1

 1。2。1 图像去雾1

 1。2。2 视频去雾2

 1。3 主要研究内容和章节安排 2

2雾霾图像的形成原理4

 2。1雾霾图像形成模型4 

 2。1。1入射光衰减模型4 

 2。1。2 大气散射模型 4 

 2。2基于物理模型的去雾算法 6 

2。2。1基于偏振信息图像去雾 6

2。2。2基于景深信息的去雾 6

2。2。3基于先验信息的雾 7 (责任编辑：qin)