基于梯度和小波变换的水下距离选通图像去噪

许廷发; 苏畅; 罗璇; 卞紫阳

doi:10.3788/CO.20160903.0301

基于梯度和小波变换的水下距离选通图像去噪

doi: 10.3788/CO.20160903.0301

北京理工大学光电学院光电成像技术与系统教育部重点实验室, 北京 100081

基金项目:

国家国际科技合作专项资助项目 No.2014DFR1096

详细信息

通讯作者:
许廷发(1968-),男,黑龙江肇东人,博士,教授,博士生导师,1992年、2000年于东北师范大学分别获得学士、硕士学位,2004年于中国科学院长春光学精密机械与研究所获得博士学位,2006年于华南理工大学电子与信息学院完成博士后工作,主要从事光电成像探测与识别等方面的研究。E-mail:xutingfa@163.com

中图分类号: TP751.1
计量
- 文章访问数: 1917
- HTML全文浏览量: 477
- PDF下载量: 934
- 被引次数: 0
出版历程
- 收稿日期: 2016-01-22
- 录用日期: 2016-02-28
- 刊出日期: 2016-01-25

Underwater range-gated image denoising based on gradient and wavelet transform

Key Laboratory of Photoelectronic Imaging Technology and System of Ministry of Education, School of Optoelectronics, Beijing Institute of Technology, Beijing 100081, China

Funds:

Supported by International S&T Cooperation Program of China No.2014DFR1096

摘要

摘要: 水体的散射效应、激光光斑、成像器件的非理想化等因素使得图像出现大量无规律粒状噪声,它们增加了水下距离选通图像的背景噪声,模糊了目标轮廓,掩盖了目标细节,降低了图像的信噪比。针对上述问题本文提出了一种基于梯度和小波变换的去噪方法。首先对图像进行余弦小波变换,得到不同频率空间的图像集。低频空间引入新的图像梯度强化方法以提高图像的纹理信息量;对应非均匀性条带的LH或HL空间做曲面拟合处理以消除非均匀性条带的影响;在HH空间去噪过程中,低层空间做非局部均值处理以保留图像相似信息,高层空间做分数阶积分处理以保留图像细节信息。最后小波逆变换得到结果图像。从实验水槽中采集水下图像进行算法验证,将改进方法与已有算法比对分析。实验表明,本文所研究的水下去噪算法,能够平滑噪声且更大限度地保留图像细节纹理,在客观评价指标上提升了6%。
- 水下图像 /
- 距离选通 /
- 图像去噪 /
- 梯度变换 /
- 小波变换
Abstract: For the scattering effect of water, the laser spot, and other non-ideal imaging device, the image appears a large number of irregular granular noise. All of them increase the background noise of underwater range-gated images, blurring the target profile, obscuring details of the target, and reducing SNR. A denoising method based on gradient and wavelet transform is proposed. Firstly, the cosine wavelet transform is used to decompose the noisy image into many different frequency space image sets. For low frequency space image, a new image gradient enhancement method is used to improve the whole image's texture information. The LH or HL space images which have the information of non-uniform strips use the surface fitting method to eliminate the whole image's non-uniform strips. In the HH space denoising process, for the lower level space images, the non-local means method is used to preserve the whole image's similarity information, and for the upper space images, the fractional integral method is used to preserve the whole image's more details. Finally, the inverse wavelet transform is used to obtain the final image. Some contrast experiment are taken using underwater images from the long sink. The results show that the denoise method proposed in this paper can smooth the noise and preserve more texture of the image at the same time that comparing with other contrast methods. The objective evaluating index is improved by 6%.
- underwater imaging /
- range-gated imaging /
- image denoising /
- gradient transform /
- wavelet transform

HTML全文

图 1 水下距离选通成像原理

Figure 1. Principle schematic of the underwater range-gated imaging

下载: 全尺寸图片幻灯片

图 2 曲面拟合示意图

Figure 2. Schematic view of surface fitting

下载: 全尺寸图片幻灯片

图 3 非局部均值滤波原理

Figure 3. Principle of the non-local mean filter

下载: 全尺寸图片幻灯片

图 4 分数阶积分模板

Figure 4. Temple of the fractional order integral

下载: 全尺寸图片幻灯片

图 5 离散余弦小波变换空间示意图

Figure 5. Space schematic corresponding to the discrete cosine wavelet transform

下载: 全尺寸图片幻灯片

图 6 本文提出的算法原理图

Figure 6. Principle schematic of algorithm proposed in this paper

下载: 全尺寸图片幻灯片

图 7 PSNR对比图

Figure 7. Contrast diagram of PSNR

下载: 全尺寸图片幻灯片

图 8 UIQI对比图

Figure 8. Contrast diagram of UIQI

下载: 全尺寸图片幻灯片

图 9 去噪实验效果图

Figure 9. Results of the denoising experiments

下载: 全尺寸图片幻灯片

图 10 去噪实验图像三维网格图

Figure 10. Three-dimensional grid map of denoising experiment images

下载: 全尺寸图片幻灯片

表 1 实验各图的主观评价表

Table 1. Subjective evaluation of test images

图像编号	图像名称	去噪效果	细节/纹理保留
a	原始图	9	9
b	噪声图	1	1
c	曲面拟合	8	3
d	非局部均值	6	6
e	分数阶积分	6	5
f	双边滤波	5	6
g	三维块匹配	7	7
h	本文方法	7	8

下载: 导出CSV

表 2 去噪方法效果评价指标

Table 2. Effect value comparation of these denoising methods

	PSNR	UIQI	耗时/s
噪声图	17.686 1	0.650 0	-
方法A	22.483 3	0.892 6	26.834 8
方法B	23.389 0	0.891 8	33.964 4
方法C	20.785 1	0.827 9	3.789 2
方法D	19.858 4	0.796 3	14.974 8
方法E	25.120 2	0.918 3	21.890 7
方法F	26.261 3	0.946 7	9.566 7
注：方法A 快速最小误差曲面拟合图像去噪
方法B 非局部均值图像去噪
方法C 分数阶积分图像去噪
方法D 双边滤波图像去噪
方法E 三维块匹配图像去噪
方法F 本文图像去噪处理方法

下载: 导出CSV

参考文献(19)

[1]	KOCAF D M,DALGLEISH F R,CAIMI F M,et al. A focus on recent developments and trends in underwater imaging[J]. Marine Technology Society Journal,2008,42(1):52-67.
[2]	宋建中.图像处理智能化的发展趋势[J].中国光学,2011,4(5):431-441. SONG J ZH. Development trend of image processing intelligence[J]. Chine Optics,2011,4(5):431-441.(in Chinese)
[3]	KANAEV A V,HOU W,RESTAINO S R. Correction methods for underwater turbulence degraded imaging[J]. SPIE,2014,9242:92421P-9.
[4]	吴海滨,刘超,于晓洋,等.视频监控图像的运动模糊方向估计[J].液晶与显示,2014,29(4):580-585. WU H B,LIU CH,YU X Y,et al. Motion blurred direction estimation for video monitor image[J]. Chinese J. Liquid Crystals and Displays,2014,29(4):580-585.(in Chinese)
[5]	朱瑞飞,魏群,王超,等.基于随机点扩散函数的多帧湍流退化图像自适应复原方法[J].中国光学,2015,8(3):368-377. ZHU R F,WEI Q,WANG CH,et al. Adaptive restoration method of multi-frame turbulence-degraded images based on stochastic point spread function[J]. Chine Optics,2015,8(3):368-377.(in Chinese)
[6]	孙玉胜,白克.基于小波变换与加权滤波的电机红外图像增强[J].液晶与显示,2010,25(3):439-443. SUN Y SH,BAI K. Enhancement of motor infrared image based on wavelet transform and weighted filtering[J]. Chinese J. Liquid Crystals and Displays,2010,25(3):439-443.(in Chinese)
[7]	王静轩,尹传历.基于DSP和FPGA的嵌入式实时图像增强系统[J].液晶与显示,2013,28(3):459-463. WANG J X,YIN CH L. Embedded color image enhancement system based on DSP and FPGA[J]. Chinese J. Liquid Crystals and Displays,2013,28(3):459-463.(in Chinese)
[8]	FREEMAN K,REICHER M,FREEMAN K,et al. A novel approach for image denoising using digital filters[J]. International Global J. Engineering Research(IGJER),2015:21-24.
[9]	YAN R,SHAO L,LIU Y. Nonlocal hierarchical dictionary learning using wavelets for image denoising[J]. Image Processing, IEEE Transactions on,2013,22(12):4689-4698.
[10]	杨文波,马天玮,刘剑.非局部变分修复法去除高密度椒盐噪声[J].中国光学,2013,6(6):876-884. YANG W B,MA T W,LIU J. Elimination of impulse noise by non-local variation inpainting method[J]. Chinese Optics,2013,6(6):876-884.(in Chinese)
[11]	刘希佳,陈宇,王文生,等.小目标识别的小波阈值去噪方法[J].中国光学,2012,5(3):248-256. LIU X J,CHEN Y,WANG W SH. De-noising algorithm of wavelet threshold for small target detection[J]. Chinese Optics,2012,5(3):248-256.(in Chinese)
[12]	杨利红,赵变红,张星祥,等.点扩散函数高斯拟合估计与遥感图像恢复[J].中国光学,2012,5(2):181-188. YANG L H,ZHAO B H,ZHANG X X,et al. Gaussian fitted estimation of point spread function and remote sensing image restoration[J]. Chinese Optics,2012,5(2):181-188.(in Chinese)
[13]	任文琦,王元全.基于梯度矢量卷积场的四阶各向异性扩散及图像去噪[J].光学精密工程,2013,21(10): 2713-2719. REN W Q,WANG Y Q. GVC-based fourth-order anisotropic diffusion for image denoising[J]. Opt. Precision Eng.,2013,21(10):2713-2719.(in Chinese)
[14]	刘巧红,李斌,林敏.结合复方向滤波器组高斯尺度混合模型及非局部均值滤波的图像去噪[J].光学精密工程,2014,22(10):2806-2814. LIU Q H,LI B,LIN M. Image denoising with dual-directional filter bank GSM model and non-local mean filter[J]. Opt. Precision Eng.,2014,22(10):2806-2814.(in Chinese)
[15]	SHAO L,YAN R,LI X,et al. From heuristic optimization to dictionary learning:a review and comprehensive comparison of image denoising algorithms[J]. IEEE,2014,44(7):1001-1013.
[16]	BUADES A,COLL B,MOREL J. A non-local algorithm for image denosing[J]. IEEE,2005:60-65.
[17]	DEKEL T,MICHAELI T,IRANI M,et al. Revealing and modifying non-local variations in a single image[J]. ACM Transactions on Graphics(TOG),2015,34(6):227.
[18]	PU Y F,ZHOU J L,YUAN X. Fractional differential mask:afractional differential based approach for multiscale texture enhancement[J]. IEEE Trans. on Image Processing,2010,19(2):491-511.
[19]	DABOV K,FOI A,KATKOVNIK V,et al. Image denoising by sparse 3-D transform-domain collaborative filtering[J]. Image Processing, IEEE Transactions on,2007,16(8):2080-2095.