一种消除高密度椒盐噪声的迭代中值滤波算法

兰霞; 刘欣鑫; 沈焕锋; 袁强强; 张良培

doi:10.13203/j.whugis20150520

一种消除高密度椒盐噪声的迭代中值滤波算法

兰霞^1,,
刘欣鑫²,
沈焕锋^{2, 3, ,},
袁强强^{3, 4},
张良培^{3, 5}

1.
中国电子科技集团公司第十研究所, 四川成都, 610036
2.
武汉大学资源与环境科学学院, 湖北武汉, 430079
3.
地球空间信息技术协同创新中心, 湖北武汉, 430079
4.
武汉大学测绘学院, 湖北武汉, 430079
5.
武汉大学测绘遥感信息工程国家重点实验室, 湖北武汉, 430079

基金项目:

国家自然科学基金 41401383

国家自然科学基金 41401396

详细信息

作者简介:
兰霞, 博士, 主要从事信号与信息处理研究。lanxia2004@163.com

通讯作者:
沈焕锋, 博士, 教授。shenhf@whu.edu.cn

中图分类号: P237.3;TP751
计量
- 文章访问数: 1892
- HTML全文浏览量: 208
- PDF下载量: 406
出版历程
- 收稿日期: 2016-07-26
- 发布日期: 2017-12-04

A Novel Median Filter to Iteratively Remove Salt-and-Pepper Noise from Highly Corrupted Images

LAN Xia^1,,
LIU Xinxin²,
SHEN Huanfeng^{2, 3, ,},
YUAN Qiangqiang^{3, 4},
ZHANG Liangpei^{3, 5}

1.
The 10 th Institute of China Electronics Technology Group Corporation, Chengdu 610036, China
2.
School of Resource and Environmental Sciences, Wuhan University, Wuhan 430079, China
3.
Collaborative Innovation Center of Geospatial Technology, Wuhan University, Wuhan 430079, China
4.
School of Geodesy and Geomatics, Wuhan University, Wuhan 430079, China
5.
The State Key Laboratory of Information Engineering in Surveying, Mapping and Remote Sensing, Wuhan University, Wuhan 430079, China

Funds:

The National Natural Science Foundation of China 41401383

The National Natural Science Foundation of China 41401396

More Information

Author Bio:
LAN Xia, PhD, specializes in signal and information processing. E-mail: lanxia2004@163.com

Corresponding author:
SHEN Huanfeng, PhD, professor. E-mail: shenhf@whu.edu.cn

摘要

摘要: 针对现有滤波方法处理高密度椒盐噪声的不足，提出一种简单有效的迭代中值滤波算法。该方法首先依据像素的强度值判断噪声点的位置，然后在循环迭代的处理框架内，对噪声像元进行逐步恢复。若噪声影像中包含足够的健康信息，则利用局部灰度差异控制项，完成对滤波结果的进一步优化。基于标准测试影像的实验表明，该方法能更为精确地恢复出被椒盐噪声污染的影像细节信息，其处理结果在目视及定量评价上均优于4种对比的滤波方法；且该方法的处理优势在影像椒盐噪声比例高达95%的情况下也依旧显著。
- 椒盐噪声 /
- 中值滤波 /
- 迭代 /
- 噪声去除 /
- 高密度
Abstract: In this paper, we propose a simple but efficient filter to effectively remove salt-and-pepper noise from highly corrupted images inspired by the corresponding limitations of existing filtering methods. After ensuring the location of ill pixels based on their intensity value, our method then utilizes the iterative processing framework to gradually restore the noisy images. When the useful information of one corrupted image is much enough, the proposed method can refine the results through particular designed criterion. The experiments from standard test images show that the proposed method can better recover the detail information and maintain the optimal performances qualitatively and quantitatively in the comparisons. Even the ratio of salt-and-pepper noise is as high as 95%, the advantage of our filter is still significant.
- salt-and-pepper noise /
- median filter /
- iteration /
- noise removal /
- high density

HTML全文

图 1 DPIMF处理流程

Figure 1. Flowchart of DPIMF

下载: 全尺寸图片幻灯片

图 2 测试影像

Figure 2. Test Images

下载: 全尺寸图片幻灯片

图 3 Aerial影像实验结果对比

Figure 3. The Restoration Results of Different Filters in Aerial

下载: 全尺寸图片幻灯片

图 4 Aerial影像实验结果放大对比

Figure 4. Detailed Regions of Restoration Results

下载: 全尺寸图片幻灯片

图 5 Boat影像实验结果对比

Figure 5. The Restoration Results of Different Filters in Boat

下载: 全尺寸图片幻灯片

图 6 Zelda影像实验结果对比

Figure 6. The Restoration Results of Different Filters in Zelda

下载: 全尺寸图片幻灯片

表 1 影像在不同噪声密度下的PSNR定量评价结果

Table 1 The Quantitative Evaluation Results Using PSNR Index Under Different Noise Condition

影像	方法	10%	40%	60%	70%	80%	90%	95%
	AMF	34.44	26.14	22.94	21.49	19.92	17.83	16.57
	DBA	34.55	26.60	22.68	20.92	18.96	16.32	15.21
Aerial	SAMF	34.56	27.01	24.49	23.14	21.74	19.88	18.58
Aerial	TIMF	36.35	28.02	24.03	22.20	20.10	17.82	16.58
	DPIMF	36.53	29.21	26.06	24.59	22.91	20.55	18.80
	AMF	37.46	28.83	25.39	24.11	22.57	20.26	18.81
	DBA	37.26	29.32	25.27	23.30	21.42	18.39	16.45
Boat	SAMF	37.48	29.95	27.25	25.89	24.51	22.64	21.31
Boat	TIMF	38.95	30.97	27.04	25.05	22.54	19.16	16.49
	DPIMF	38.97	31.75	28.60	26.97	25.42	23.28	21.52
	AMF	40.41	31.56	28.10	26.47	24.43	21.57	19.45
	DBA	41.12	32.19	27.92	25.07	22.57	18.21	15.84
Peppers	SAMF	42.12	33.87	31.00	29.56	27.63	24.97	22.97
Peppers	TIMF	42.78	34.17	29.91	26.93	23.58	18.50	15.14
	DPIMF	42.94	35.46	32.26	30.07	28.32	25.69	23.60
	AMF	44.48	35.65	32.18	30.17	28.58	25.86	23.30
	DBA	44.72	36.08	31.27	28.91	26.10	22.21	19.24
Zelda	SAMF	44.77	36.94	34.32	32.78	31.12	28.95	26.97
Zelda	TIMF	46.07	37.77	33.59	31.06	27.57	22.52	18.37
	DPIMF	46.05	38.66	35.26	33.93	32.15	29.76	27.66

下载: 导出CSV

表 2 影像在不同噪声下密度的MAE定量评价结果

Table 2 The Quantitative Evaluation Results Using MAE Index Under Different Noise Condition

影像	方法	10%	40%	60%	70%	80%	90%	95%
	AMF	0.86	4.47	7.98	10.26	13.34	18.46	22.73
	DBA	0.86	4.23	8.16	10.92	15.04	22.69	28.22
Aerial	SAMF	0.84	4.13	6.80	8.65	11.02	14.63	17.77
Aerial	TIMF	0.69	3.63	7.07	9.57	13.45	20.09	25.24
	DPIMF	0.68	3.16	5.62	7.17	9.32	13.07	16.85
	AMF	0.65	3.35	5.93	7.50	9.60	13.25	16.55
	DBA	0.69	3.18	5.97	8.07	10.89	16.76	22.98
Boat	SAMF	0.64	2.98	4.89	6.16	7.77	10.22	12.36
Boat	TIMF	0.56	2.74	5.12	6.97	10.02	16.78	25.96
	DPIMF	0.56	2.53	4.28	5.56	6.99	9.42	11.86
	AMF	0.46	2.32	4.08	5.25	6.98	10.17	13.59
	DBA	0.44	2.16	4.03	5.75	8.30	15.18	22.66
Peppers	SAMF	0.39	1.85	3.05	3.81	4.96	6.93	9.12
Peppers	TIMF	0.38	1.83	3.38	4.91	7.73	16.29	29.18
	DPIMF	0.38	1.68	2.82	3.77	4.80	6.61	8.51
	AMF	0.33	1.70	3.03	3.94	5.18	7.47	10.18
	DBA	0.33	1.60	3.10	4.28	6.30	10.84	16.95
Zelda	SAMF	0.30	1.43	2.36	3.00	3.84	5.27	6.87
Zelda	TIMF	0.28	1.35	2.55	3.62	5.78	11.99	22.21
	DPIMF	0.28	1.27	2.22	2.78	3.55	4.87	6.28

下载: 导出CSV

表 3 Zelda影像不同噪声密度下不同方法的计算时间比较/s

Table 3 Running Time of the Different Methods Under Different Noise Condition in Zelda/s

方法	10%	40%	70%	95%
AMF	1.23	4.16	10.37	46.61
DBA	2.58	2.49	2.57	2.63
SAMF	0.64	2.06	2.85	6.48
TIMF	0.31	1.01	1.71	2.32
DPIMF	1.58	6.16	1.73	3.82

下载: 导出CSV

参考文献(16)

[1]	Bovik A. Handbook of Image and Processing[M]. New York:Academic Press, 2000
[2]	Rosenfeld A, Woods R E. Digital Picture Processing[M]. New York:Academic Press, 1982
[3]	Nodes T, Gallagher N C. The Output Distribution of Median Type Filters[J]. IEEE Transactions on Communications, 1984, 32(5):532-541 doi: 10.1109/TCOM.1984.1096099
[4]	Brownrigg D. The Weighted Median Filter[J]. Communications of the ACM, 1984, 27(8):807-818 doi: 10.1145/358198.358222
[5]	Ko S J, Lee Y H. Center Weighted Median Filters and Their Applications to Image Enhancement[J]. IEEE Transactions on Circuits & Systems, 1991, 38(9):984-993 http://ieeexplore.ieee.org/xpl/articleDetails.jsp?arnumber=83870&contentType=Journals+%26+Magazines&punumber%3D31
[6]	Abreu E, Lightstone M, Mitra S K, et al. A New Efficient Approach for the Removal of Impulse Noise from Highly Corrupted Images[J]. IEEE Transactions on Image Processing, 1996, 5(6):1012-1025 doi: 10.1109/83.503916
[7]	Sun T, Neuvo Y. Detail-Preserving Median Based Filters in Image Processing[J]. Pattern Recognition Letters, 1994, 15(4):341-347 doi: 10.1016/0167-8655(94)90082-5
[8]	李树涛, 王耀南.图象椒盐噪声的非线性自适应滤除[J].中国图象图形学报, 2000, 5(12):999-1001 doi: 10.3969/j.issn.1006-8961.2000.12.004 Li Shutao, Wang Yaonan. Non-Linear Adaptive Removal of Salt and Pepper Noise from Images[J]. Journal of Image and Graphics, 2000, 5(12):999-1001 doi: 10.3969/j.issn.1006-8961.2000.12.004
[9]	邢藏菊, 王守觉, 邓浩江, 等.一种基于极值中值的新型滤波算法[J].中国图象图形学报, 2001, 6(6):533-536 http://www.cqvip.com/QK/90287X/2001006/5376853.html Xing Cangju, Wang Shoujue, Deng Haojiang, et al. A New Filtering Algorithm Based on Extremum and Median Value[J]. Journal of Image and Graphics, 2001, 6(6):533-536 http://www.cqvip.com/QK/90287X/2001006/5376853.html
[10]	Zhang S, Karim M A. A New Impulse Noise Detector for Switching Median Filters[J]. IEEE Signal Processing Letters, 2002, 9(11):360-363 doi: 10.1109/LSP.2002.805310
[11]	Chan R H, Ho C W, Nikolova M. Salt-and-Pepper Noise Removal by Median-Type Noise Detectors and Detail Preserving Regularization[J]. IEEE Transactions on Signal Processing, 2005, 14(10):1479-1485 http://ci.nii.ac.jp/naid/80017549968
[12]	Srinivasan K S, Ebenezer D. A New Fast and Efficient Decision-Based on Algorithm for Removal of High-Density Impulse Noise[J]. IEEE Signal Processing Letters, 2007, 14(3):189-192 doi: 10.1109/LSP.2006.884018
[13]	Ibrahim H, Kong N S P, Ng T F. Simple Adaptive Median Filter for the Removal of Impulse Noise from Highly Corrupted Images[J]. IEEE Transactions on Consumer Electronics, 2008, 54(4):1920-1927 doi: 10.1109/TCE.2008.4711254
[14]	Wang Z, Zhang D. Progressive Switching Median Filter for the Removal Impulse Noise from Highly Corrupted Images[J]. IEEE Transactions on Circuits and Systems Ⅱ Analog & Digital Signal Processing, 1999, 46(1):78-80 http://citeseerx.ist.psu.edu/showciting?cid=260222
[15]	Ahmed F, Das S. Removal of High-Density Salt-and-Pepper Noise in Images with an Iterative Adaptive Fuzzy Filter Using Alpha-Trimmed Mean[J]. IEEE Transactions on Fuzzy Systems, 2014, 22(5):1352-1358 doi: 10.1109/TFUZZ.2013.2286634
[16]	Bhadouria V S, Ghoshal D, Siddiqi A H. A New Approach for High Density Saturated Impulse Noise Removal Using Decision-Based Coupled Window Median Filter[J]. Signal Image and Video Processing, 2014, 8(1):71-84 doi: 10.1007/s11760-013-0487-5

施引文献(7)

期刊类型引用(4)

1.	朱杰，郑加柱，陈红华，杨静，胡平昌，陆敏燕. 结合POI数据的南京市商业中心识别与集聚特征研究. 现代测绘. 2022(06): 34-39 . 百度学术
2.	金澄，安晓亚，陈占龙，马啸川. 矢量居民地多边形多级图划分聚类方法. 武汉大学学报(信息科学版). 2021(01): 19-29 . 百度学术
3.	张铭龙，何贞铭. 基于因子分析法的城市商业中心抽取研究. 地理空间信息. 2021(08): 58-60+64+5 . 百度学术
4.	李卫东，张铭龙，段金龙. 基于POI数据的南京市空间格局定量研究. 世界地理研究. 2020(02): 317-326 . 百度学术