公共安全低质图像质量提升算法与软件开发
（1）研究非局部去噪的原理，实现相应的去噪算法，将其与双边滤波和 BM3D算法比较PSNR和SSIM，同时比较去噪处理时间。实验表明，非局部去噪效果略逊与 BM3D，但处理时间较短。
（2）实现了暗通道去雾算法，并优化相关参数，获得较优的去雾效果。 26049
（3）研究了参数化非盲和稀疏正则化盲的模糊核估计方法，并针对估计出的模糊核分别使用了文纳滤波和 TV 正则来去模糊。比较两种方法的去模糊效果，实验证明 TV 正则化去模糊有较高的 PSNR和SSIM。
（4）采用了基于插件式的软件开发方式，给出了一个使用 C#编写的图像质量提升软件。
毕业论文关键词 非局部去噪 暗通道去雾 运动模糊 插件式
Title Algorithms and Software Development for Improving Low-quality Images Acquired from Public Security Area
Abstract
Due to the bad environment and machine problems, images and videos acquired
from public security area could be noisy, hazy or blurry. Thus, it would be a
great help to police in identifying key information if proper improving image
algorithms are applied. Focusing on the three factors that cause images in a
low quality, algorithms are studied and a software is developed. The main work
are follows:
(1)Non-local means and NLM-P algorithm are studied. NLM-P is compared with
bilateral filtering and BM3D in PSNR, SSIM and CPU-time. Experimental results
show that NLM-P is not as good as BM3D in denosing, but NLM-P runs faster.
(2)Dark channel prior algorithm is implemented, with parameters adjusted to
make foggy images clearer.
(3)Parametric non-blind and sparse regularized blind ker estimation are
studied. Wiener filtering and TV regularized method are applied to deblur.
Experimental results show that TV regularized method has a higher PSNR and
SSIM.
(4)These algorithms are programmed using the plug-in technology. An image
improvement software developed by C# is offered.
Keywords NLM-P dark channel motion deblur plug-in
目 次
1 绪论 1
1.1 研究的目的与意义 1
1.2 图像质量提升的国内外发展现状 1
1.3 图像的质量评价 3
1.4 本文的主要内容 4
2 非局部无参逐像素去噪算法 5
2.1 非局部方法原理 5
2.2 非局部去噪算法（NLM-P）与改进 6
2.3 NLM-P 与双边滤波、BM3D 的比较 8
2.4 本章小结 13
3 基于暗通道先验的去雾算法 14
3.1 暗通道去雾原理 14
3.2 参数设置对去雾效果的影响15
3.4 本章小结 17
4 模糊核参数化非盲估计和稀疏正则化盲估计的运动模糊算法 18
4.1 模糊核参数化非盲估计和文纳滤波 18
4.2 基于梯度�
0稀疏的模糊核估计与 TV 正则化去模糊 20
4.3 两种去模糊方法的比较 23
4.4 本章小结 25
5 基于插件式的图像质量提升软件 26
5.1 基于插件式的 WPF 软件开发 26