重加权总变分结合hyper-Laplacian的图像盲复原方法 下载: 890次
Blind Image Restoration Method Based on Reweighted Graph Total Variation and Hyper-Laplacian
图 & 表
图 1. 本文盲复原算法的流程
Fig. 1. Process of the proposed blind restoration algorithm
下载图片 查看原文
图 2. 三种图像块。(a)清晰的自然图像块;(b)模糊图像块;(c)潜在中间图像块
Fig. 2. Three image blocks. (a) Clear natural image block; (b) blurred image block; (c) intermediate latent image block
下载图片 查看原文
图 3. 图2中三种图像块的权重分布。(a)清晰自然图像;(b)模糊图像;(c)潜在中间图像
Fig. 3. Weight distribution of three images in Fig. 2. (a) Clear natural image; (b) blurred image; (c) intermediate latent image
下载图片 查看原文
图 4. 模糊图像blue car盲复原效果对比。(a)清晰图像;(b)模糊图像;(c)文献[
9]算法的复原效果;(d)文献[
14]算法的复原效果;(e)本文算法的复原效果
Fig. 4. Comparison of blind restoration effect of blurred image of blue car. (a) Clear image; (b) blurred image; (c) restoration effect by method in Ref.[9]; (d) restoration effect by method in Ref.[14]; (e) restoration effect by proposed method
下载图片 查看原文
图 5. 模糊图像red car的盲复原效果对比。(a)模糊图像;(b)文献[
9]算法的复原效果;(c)文献[
14]算法的复原效果;(d)本文算法的复原效果;(e)潜在中间图像
Fig. 5. Comparison of blind restoration effect of blurred image of red car. (a) Blurred image; (b) restoration effect by method in Ref.[9]; (c) restoration effect by method in Ref.[14]; (d) restoration effect by proposed method; (e) intermediate latent image
下载图片 查看原文
图 6. 模糊图像monkey的盲复原效果对比。(a)模糊图像;(b)文献[
9]算法的复原效果;(c)文献[
14]算法的复原效果;(d)本文算法的复原效果;(e)潜在中间图像
Fig. 6. Comparison of blind restoration effect of blurred image of monkey. (a) Blurred image; (b) restoration effect by method in Ref.[9]; (c) restoration effect by method in Ref.[14]; (d) restoration effect by proposed method; (e) intermediate latent image
下载图片 查看原文
图 7. 模糊图像summer house的盲复原对比效果。(a)模糊图像;(b)文献[
9]算法的复原效果;(c)文献[
14]算法的复原效果;(d)本文算法的复原效果;(e)潜在中间图像
Fig. 7. Comparison of blind restoration effect of blurred image of summer house. (a) Blurred image; (b) restoration effect by method in Ref.[9]; (c) restoration effect by method in Ref.[14]; (d) restoration effect by proposed method; (e) intermediate latent image
下载图片 查看原文
图 8. 图7(a)复原前、后的边缘权重分布图。(a)模糊图像的权重分布;(b)潜在中间图像的权重分布;(c)本文算法复原图像的权重分布
Fig. 8. Weight distribution of Fig. 7(a) before and after restoration. (a) Weight distribution of blurred image; (b) Weight distribution of intermediate latent image; (c) Weight distribution of restored image by proposed method
下载图片 查看原文
表 1图像red car的无参考质量盲评价指标
Table1. Blind evaluation index without reference quality of image of red car
Method | Informationentropy | Standarddeviation | Mean value |
---|
Blurred image | 7.3245 | 2.1501×103 | 90.9321 | Method in Ref.[9] | 7.5822 | 2.7107×103 | 89.4635 | Method in Ref.[14] | 7.3393 | 2.6849×103 | 91.3737 | Proposed method | 7.4875 | 2.8116×103 | 91.4632 |
|
查看原文
表 2图像monkey的无参考质量盲评价指标
Table2. Blind evaluation index without reference quality of image of monkey
Parameter | Informationentropy | Standarddeviation | Mean value |
---|
Blurred image | 7.0363 | 945.4009 | 107.0265 | Method in Ref.[9] | 7.1859 | 1.566×103 | 106.0104 | Method in Ref.[14] | 7.1628 | 1.1663×103 | 107.9668 | Proposed method | 7.2325 | 1.2691×103 | 108.6190 |
|
查看原文
表 3图像summer house的无参考质量盲评价指标
Table3. Blind evaluation index without reference quality of image of summer house
Parameter | Informationentropy | Standarddeviation | Mean value |
---|
Blurred image | 7.4303 | 3.3740×103 | 109.9294 | Method in Ref.[9] | 7.5492 | 3.4198×103 | 110.4320 | Method in Ref.[14] | 7.5213 | 3.6317×103 | 110.4827 | Proposed method | 7.6903 | 3.8078×103 | 110.6824 |
|
查看原文
许泽海, 宋海燕. 重加权总变分结合hyper-Laplacian的图像盲复原方法[J]. 激光与光电子学进展, 2020, 57(8): 081025. Zehai Xu, Haiyan Song. Blind Image Restoration Method Based on Reweighted Graph Total Variation and Hyper-Laplacian[J]. Laser & Optoelectronics Progress, 2020, 57(8): 081025.