基于非局部先验的单幅图像去雾算法 下载: 1220次
Single Image Dehazing Algorithm Based On Non-Local Prior
空军工程大学航空航天工程学院, 陕西 西安 710038
图 & 表
图 1. 非局部图像去雾。(a)清晰无雾图像及其聚类成的簇;(b)有雾图像及其聚类成的雾线
Fig. 1. Non-local image dehazing. (a) Haze-free image and corresponding clusters; (b) hazy image and corresponding haze-lines
下载图片 查看原文
图 2. 雾线。(a)有雾图像;(b)原始图像中标记的像素点在RGB颜色空间中形成的雾线
Fig. 2. Haze-lines. (a) Haze image; (b) pixels marked in the original image formed a haze-line in RGB color space
下载图片 查看原文
图 3. 初始估计透射率。(a) 雾天图像;(b) r(x);(c) r^max(x);(d) t~(x)
Fig. 3. Estimation of initial transmission. (a) Haze image; (b) r(x); (c) r^max(x); (d) t~(x)
下载图片 查看原文
图 4. 现有算法的去雾结果。(a)原始有雾图像; (b)文献[
11]算法的去雾图像; (c)文献[
13] 算法的去雾图像
Fig. 4. Dehazing results using existing algorithms. (a) Initial haze image; (b) dehazing image of Ref. [11] method; (c) dehazing image of Ref. [13] method
下载图片 查看原文
图 5. 透射率优化及去雾结果对比。(a)雾天图像; (b)文献[
10]算法; (c)文献[
9]算法; (d)本文算法
Fig. 5. Comparison of the transmission optimization and the dehazing results. (a) Haze image; (b) Ref. [10] algorithm; (c) Ref. [9] algorithm; (d) proposed algorithm
下载图片 查看原文
图 6. 不同算法的去雾结果。(a)雾天图像; (b)文献[
10]算法; (c)文献[
11]算法;(d)文献[
9]算法; (e)文献[
13]算法; (f)本文算法
Fig. 6. Dehazing results using different algorithms.(a) Haze images; (b) Ref. [10] algorithm; (c) Ref. [11] algorithm; (d) Ref. [9] algorithm; (e) Ref. [13] algorithm; (f) proposed algorithm
下载图片 查看原文
图 7. 边缘区域对比。(a)雾天图像;(b)文献[
10]算法;(c)文献[
13]算法;(d)文献[
27]算法;(e)本文算法
Fig. 7. Comparison of the edge regions. (a) Haze image; (b) Ref. [10] algorithm; (c) Ref. [13] algorithm; (d) Ref. [27] algorithm; (e) proposed algorithm
下载图片 查看原文
表 1图6中各图的信息熵
Table1. Entropy of the images in Fig. 6
Image | Input image | Ref. [10] algorithm | Ref. [9] algorithm | Ref. [11] algorithm | Ref. [13] algorithm | Proposed algorithm |
---|
1 | 7.4180 | 7.3373 | 5.8372 | 7.4390 | 7.3429 | 7.5031 | 2 | 7.4594 | 7.5374 | 4.5518 | 7.3628 | 7.5539 | 7.6932 | 3 | 7.0615 | 7.6721 | 4.4150 | 7.2419 | 7.6721 | 7.6721 | 4 | 7.6660 | 6.8968 | 4.7989 | 7.1874 | 7.0614 | 7.9609 | 5 | 7.6423 | 7.7042 | 5.0373 | 7.5910 | 7.6629 | 7.7446 |
|
查看原文
表 2图6中各图的PSNR
Table2. PSNR of the images in Fig. 6
Image | Ref. [10] algorithm | Ref. [9] algorithm | Ref. [11] algorithm | Ref. [13] algorithm | Proposed algorithm |
---|
1 | 17.0512 | 9.2792 | 16.8589 | 17.1095 | 18.6577 | 2 | 21.8279 | 7.9024 | 12.8265 | 22.0113 | 24.1207 | 3 | 18.0434 | 7.2664 | 11.5222 | 18.0434 | 18.2991 | 4 | 13.5857 | 9.0701 | 14.1681 | 14.4161 | 14.6647 | 5 | 16.6095 | 8.5369 | 14.3926 | 16.2795 | 17.5110 |
|
查看原文
表 3不同算法的运行时间对比
Table3. Comparison of running time with different algorithmss
Image | Ref. [10] algorithm | Ref. [9] algorithm | Ref. [11] algorithm | Ref. [13] algorithm | Proposed algorithm |
---|
1 | 108.8943 | 105.2638 | 35.0985 | 10.3396 | 4.2958 | 2 | 23.8031 | 22.2016 | 30.1716 | 1.3086 | 1.0075 | 3 | 48.6528 | 40.4519 | 27.2880 | 2.9783 | 2.2884 | 4 | 53.5039 | 48.241 | 39.7485 | 2.6732 | 2.1691 | 5 | 50.6527 | 42.5846 | 36.3287 | 2.5290 | 1.4412 |
|
查看原文
董亚运, 毕笃彦, 何林远, 马时平. 基于非局部先验的单幅图像去雾算法[J]. 光学学报, 2017, 37(11): 1110001. Yayun Dong, Duyan Bi, Linyuan He, Shiping Ma. Single Image Dehazing Algorithm Based On Non-Local Prior[J]. Acta Optica Sinica, 2017, 37(11): 1110001.