摘要为了消除雾霾天气对道路视频监控系统的影响，提高雾霾天气下得到的视频图像的视觉效果，我们需要对于雾霾天气下监控所拍摄到的模糊图像进行清晰化处理。通过分析对比物理模型和非物理模型这两类方法，利用Retinex算法增强恶劣天气下得到的模糊图像，并对于Retinex算法还存在的不足之处，提出相对应的优化方案，使 Retinex算法的复杂程度得到降低。本文通过实验论证了所提出的Retinex图像增强算法比传统方法更好，它可以使雾霾天气下图像显示的更加清晰。通过实验可以看到，通过处理可以使得雾霾天气下不清晰的模糊图像变得清晰，也让其色彩更加趋于自然。本篇文章从Retinex算法出发，针对图像去雾算法进行了深入研究，由Retinex算法运算效率不高造成去雾处理后图像变暗的现象，提出了用递归高斯滤波加速Retinex算法的进行，同时可以采用线性拉伸的方法使得图像的对比度较之原图有所提高。进而大幅度提升此算法的运算效率，得到比较好的去雾效果。77360

毕业论文关键词：雾霾天气  模糊图像  清晰化处理  Retinex算法

Research on the method of clear image under haze weather

Abstract  Haze weather managementlevel blurred image processing, its aim is to eliminate the fog haze weather influence on road video monitoring system, improve the haze weather get video image visual effect。 Analyzed by physical model and physical model method, the two multiscale Retinex algorithm was applied to haze image enhancement processing, and aims at its existence insufficiency, proposed the corresponding improvement program。 The algorithm makes full use of 2 d image gaussian convolution separability and multiscale continuous estimate the feasibility of the irradiation light, reducing the Retinex algorithm complexity。 The experimental results show that compared with the methods of histogram equalization and improved multiscale Retinex image enhancement algorithm can more effectively improve the sharpness of the image from haze, the processed image is enhanced details highlighted, real natural colours。 This paper deeply studied the image to fog algorithm, based on Retinex algorithm for Retinex algorithm operation inefficient and go after the fog image appear dim phenomenon, put forward by using recursive gaussian filtering Retinex algorithm for acceleration and the method of using linear stretching improve the contrast of the image。 Thus greatly improve the operation efficiency of the algorithm, to obtain good to mist effect。

Key Words: Haze weather  Blurred image  Embarrassment for processing Retinex algorithm

目  录

摘要Ⅰ

AbstractⅡ

目录Ⅲ

图清单Ⅳ

变量注释表Ⅳ

1 绪论1

1。1研究背景及意义1

1。2研究现状及问题1

1。3本文的研究工作2

2 图像增强方法概述3

2。1线性变换和分段线性变换3

2。2 图像边缘增强4

2。3 直方图均衡化5

2。4 直方图规定化6

2。5 同态滤波7

3 雾霾天气下模糊图像的清晰化处理9

3。1 颜色恒常性与Retinex理论9

3。2 基于全局特征的Retinex算法分析12

3。3 算法流程16

3。4 增强结果显示17

4 结论18

参考文献19

致谢20

图清单

图序号 图名称 页码

图21 三段式线性变换 4

图22 拉普拉斯算法3×3模板 5