一种用于极化SAR图像的快速超像素分割算法

张月; 邹焕新; 邵宁远; 周石琳; 计科峰

doi:10.12000/JR17018

一种用于极化SAR图像的快速超像素分割算法

DOI: 10.12000/JR17018

国防科学技术大学电子科学与工程学院长沙 410073

基金项目: 国家自然科学基金(61331015, 61372163)

详细信息

作者简介:
张月：张月(1990–)，女，河南人，现为国防科技大学电子科学与工程学院硕士研究生，主要研究方向为极化SAR图像地物分类。E-mail: YueZhang15a@163.com

邹焕新(1973–)，男，广东人，现任国防科技大学电子科学与工程学院副教授，硕士生导师，主要研究方向为SAR图像解译、多源遥感信息融合等。E-mail: hxzou2008@163.com

邵宁远(1995–)，女，江苏人，现为国防科技大学电子科学与工程学院硕士研究生，主要研究方向为多源遥感数据变化检测。E-mail: ningyuanshao@163.com

周石琳(1965–)，男，湖南人，现任国防科技大学电子科学与工程学院教授，博士生导师，主要研究方向为计算机视觉与智能信息处理、多源遥感信息融合等。E-mail: slzhoumail@163.com

计科峰(1974–)，男，陕西人，现任国防科技大学电子科学与工程学院副教授，硕士生导师，主要研究方向为SAR图像解译。E-mail: jikefeng@nudt.edu.cn

通讯作者:
邹焕新 hxzou2008@163.com

中图分类号: TN957
计量
- 文章访问数: 3425
- HTML全文浏览量: 934
- PDF下载量: 766
- 被引次数: 0
出版历程
- 收稿日期: 2017-02-28
- 修回日期: 2017-07-04
- 网络出版日期: 2017-10-28

Fast Superpixel Segmentation Algorithm for PolSAR Images

College of Electronic Science and Engineering, National University of Defense Technology, Changsha 410073, China

Funds: The National Natural Science Foundation of China (61331015, 61372163)

摘要

摘要: 超像素分割算法作为预处理技术应该具有如下特性：快速的运算速度、较高的边缘贴合度及同质区域规则的形状。基于迭代边缘精炼(Iterative Edge Refinement, IER)的超像素快速分割算法在光学图像上取得了较好的效果。但是，由于极化SAR图像受相干斑噪声影响，并且存在许多小块的或者细长的区域，因此，当将IER算法直接用于极化SAR图像进行超像素分割时，难以获得理想的结果。针对以上问题，该文在初始化步骤，将不稳定像素点集初始化为极化SAR图像中的所有像素点而非网格边缘像素点；在为不稳定像素点的局部重贴标签中，用快速的修正Wishart距离代替颜色空间的欧式距离；然后，采用基于不相似度的后处理算法，在移除生成的孤立小面积超像素的同时保留强散射点目标；最后，基于一幅仿真图像和一幅AirSAR实测极化SAR图像，与其他3种较优的算法进行了对比实验。实验结果表明，就几种常用评价标准而言，该文算法具有较好的特性，而且该文算法计算效率高，能够生成边缘贴合度较高的、形状规则的超像素。
- 超像素分割 /
- 边缘精炼 /
- 修正Wishart距离 /
- 不稳定像素点 /
- 极化SAR图像
Abstract: As a pre-processing technique, superpixel segmentation algorithms should be of high computational efficiency, accurate boundary adherence and regular shape in homogeneous regions. A fast superpixel segmentation algorithm based on Iterative Edge Refinement (IER) has shown to be applicable on optical images. However, it is difficult to obtain the ideal results when IER is applied directly to PolSAR images due to the speckle noise and small or slim regions in PolSAR images. To address these problems, in this study, the unstable pixel set is initialized as all the pixels in the PolSAR image instead of the initial grid edge pixels. In the local relabeling of the unstable pixels, the fast revised Wishart distance is utilized instead of the Euclidean distance in CIELAB color space. Then, a post-processing procedure based on dissimilarity measure is empolyed to remove isolated small superpixels as well as to retain the strong point targets. Finally, extensive experiments based on a simulated image and a real-world PolSAR image from Airborne Synthetic Aperture Radar (AirSAR) are conducted, showing that the proposed algorithm, compared with three state-of-the-art methods, performs better in terms of several commonly used evaluation criteria with high computational efficiency, accurate boundary adherence, and homogeneous regularity.
- Superpixel segmentation /
- Edge refinement /
- Revised Wishart distance /
- Unstable pixels /
- Polarimetric Synthetic Aperture Radar (PolSAR) image

HTML全文

图 1 两种初始化方法的示意图。C_i代表第i个聚类的中心，S为初始网格边长。黑色的像素点是初始聚类中心。IER的初始不稳定点集为黄色的像素点，而本文算法的初始不稳定点集为黄色、白色和黑色的像素点。

Figure 1. The sketch map of initialization of two methods. C_i indicates the ith cluster center, and S is the initial grid width. The pixels filled with black are the initial cluster centers. The initial unstable pixels of the IER algorithm are the yellow pixels, while the initial unstable pixels of the proposed method are the yellow, white, and black pixels.

下载: 全尺寸图片幻灯片

图 2 仿真极化SAR的Pauli_RGB图像

Figure 2. Pauli_RGB image of the simulated image

下载: 全尺寸图片幻灯片

图 5 4种算法生成的超像素。第2行是4种算法生成的超像素, 叠加到Pauli-RGB图像上的红线是超像素的边缘。第3行中每个像素点的颜色由其所属超像素内平均颜色所替代

Figure 5. Generated superpixels of the four competitive methods. The second row denotes the final superpixel maps of different methods. The red lines superimposed onto the Pauli-RGB images depict the superpixel boundaries. The third row gives the representation maps, where the color of each pixel is replaced by the average value of the superpixel to which this pixel belongs

下载: 全尺寸图片幻灯片

图 3 基于仿真图像的两种初始化算法的比较结果图

Figure 3. Comparison of two methods of initialization based on the simulated PolSAR images

下载: 全尺寸图片幻灯片

图 4 4种算法的结果图

Figure 4. The results of four algorithms

下载: 全尺寸图片幻灯片

图 6 3个极化SAR图像块(第1列)及由标准SLIC(第2列)、IER(第3列)，SLIC-GC(第4列)和本文算法(第5列)产生的相应的超像素结果图

Figure 6. Three PolSAR image patches (first column) and corresponding superpixels provided by the standard SLIC (second column), IER (third column), SLIC-GC (fourth column), and the proposed method (fifth column)

下载: 全尺寸图片幻灯片

表 1 4种算法基于AirSAR实测极化SAR图像生成超像素的时间(以s为单位)

Table 1. Running time (in seconds) of four methods for real-world AirSAR PolSAR image

算法	聚类时间(s)	后处理时间(s)	总时间(s)
标准SLIC算法	330.981	44.850	375.831
IER算法	212.264	16.591	228.855
SLIC-GC算法	3433.400	3.460	3436.860
本文算法	366.469	16.877	383.346

下载: 导出CSV

参考文献(29)

[1]	Song H, Yang W, Xu X, et al.. Unsupervised PolSAR imagery classification based on Jensen-Bregman LogDet divergence[C]. European Conference on Synthetic Aperture Radar, EUSAR, Berlin, 2014: 1–4.
[2]	孙勋, 黄平平, 涂尚坦,等. 利用多特征融合和集成学习的极化SAR图像分类[J]. 雷达学报, 2016, 5(6): 692–700. Sun Xun, Huang Pingping, Tu Shangtan, et al. Polarimetric SAR image classification using multiple-feature fusion and ensemble learning[J]. Journal of Radars, 2016, 5(6): 692–700.
[3]	Dabboor M, Collins M J, Karathanassi V, et al. An unsupervised classification approach for polarimetric SAR data based on the Chernoff distance for complex Wishart distribution[J]. IEEE Transactions on Geoscience and Remote Sensing, 2013, 51(7): 4200–4213. DOI: 10.1109/TGRS.2012.2227755.
[4]	滑文强, 王爽, 侯彪. 基于半监督学习的SVM-Wishart极化SAR图像分类方法[J]. 雷达学报, 2015, 4(2): 93–98. Hua Wenqiang, Wang Shuang, and Hou Biao. Semi-supervised learning for classification of polarimetric SAR images based on SVM-Wishart[J]. Journal of Radars, 2015, 4(2): 93–98.
[5]	Xu Q, Chen Q H, Yang S, et al. Superpixel-based classification using K distribution and spatial context for polarimetric SAR images[J]. Rmote Sensing, 2016, 8(8): 619. DOI: 10.3390/rs8080619.
[6]	Wu Y H, Ji K F, Yu W X, et al. Region-based classification of polarimetric SAR images using Wishart MRF[J]. IEEE Geoscience and Remote Sensing Letters, 2008, 5(4): 668–672. DOI: 10.1109/LGRS.2008.2002263.
[7]	Ren X and Malik J. Learning a classification model for segmentation[C]. IEEE International Conference on Computer Vision. Nice, France, 2003: 10–17.
[8]	Gong M G, Su L Z, Jia M, et al. Fuzzy clustering with a modified MRF energy function for change detection in synthetic aperture radar images[J]. IEEE Transactions on Fuzzy Systems, 2014, 22(1): 98–109. DOI: 10.1109/TFUZZ.2013.2249072.
[9]	Xie L, Zhang H, Wang C, et al.. Superpixel-based PolSAR images change detection[C]. 2015 IEEE 5th Asia-Pacific Conference on Synthetic Aperture Radar, Singapore, 2015.
[10]	Wang S, Lu H, Yang F, et al.. Superpixel tracking[C]. IEEE International Conference on Computer Vision, Barcelona, Spain, 2011: 1323–1330.
[11]	Liu B, Hu H, Wang H Y, et al. Superpixel-based classification with an adaptive number of classes for polarimetric sar images[J]. IEEE Transactions on Geoscience and Remote Sensing, 2013, 51(2): 907–924. DOI: 10.1109/TGRS.2012.2203358.
[12]	Xiang D L, Tang T, Zhao L J, et al. Superpixel generating algorithm based on pixel intensity and location similarity for SAR image classification[J]. IEEE Geoscience and Remote Sensing Letters, 2013, 10(6): 1414–1418. DOI: 10.1109/LGRS.2013.2259214.
[13]	Xing Y X, Zhang Y, Li N, et al. Improved superpixel-based polarimetric synthetic aperture radar image classification integrating color features[J]. Journal of Applied Remote Sensing, 2016, 10(2): 026026. DOI: 10.1117/1.JRS.10.026026.
[14]	Liu M Y, Tuzel O, Ramalingam S, et al.. Entropy rate superpixel segmentation[C]. IEEE Conference on Computer Vision and Pattern Recognition, Providence, 2011: 2097–2104.
[15]	Zhang Y H, Hartley R, Mashford J, et al.. Superpixels via pseudo-Boolean optimization[C]. IEEE International Conference on Computer Vision, Barcelona, 2011: 1387–1394.
[16]	Vedaldi A and Soatto S. Quick shift and kernel methods for mode seeking[C]. European Conference on Computer Vision, Berlin, 2008: 705–718.
[17]	Mester R, Conrad C, and Guevara A. Multichannel Segmentation Using Contour Relaxation: Fast Super-Pixels and Temporal Propagation[M]. Heyden A and Kahl F, eds. Image Analysis. Berlin Heidelberg: Springer, 2011.
[18]	Den Bergh M V, Boix X, Roig G, et al. SEEDS: Superpixels extracted via energy-driven sampling[J]. International Journal of Computer Vision, 2015, 111(3): 298–314. DOI: 10.1007/s11263-014-0744-2.
[19]	Achanta R, Shaji A, Smith K, et al.. SLIC superpixels[R]. EPFL, 2010.
[20]	Zou H, Qin X, Zhou S, et al. A likelihood-based SLIC superpixel algorithm for SAR images using generalized Gamma distribution[J]. Sensors, 2016, 16(7): E1107. DOI: 10.3390/s16071107.
[21]	Feng J L, Cao Z J, and Pi Y M. Polarimetric contextual classification of PolSAR images using sparse representation and superpixels[J]. Remote Sensing, 2014, 6(8): 7158–7181. DOI: 10.3390/rs6087158.
[22]	Qin F C, Guo J M, and Lang F K. Superpixel segmentation for polarimetric SAR imagery using local iterative clustering[J]. IEEE Geoscience and Remote Sensing Letters, 2015, 12(1): 13–17. DOI: 10.1109/LGRS.2014.2322960.
[23]	Zhu S, Cao D, Jiang S, et al. Fast superpixel segmentation by iterative edge refinement[J]. Electronics Letters, 2015, 51(3): 230–232. DOI: 10.1049/el.2014.3379.
[24]	Jiao L C and Liu F. Wishart deep stacking network for fast PolSAR image classification[J]. IEEE Transactions on Image Processing, 2016, 25(7): 3273–3286. DOI: 10.1109/TIP.2016.2567069.
[25]	Kersten P R, Lee J S, and Ainsworth T L. Unsupervised classification of polarimetric synthetic aperture radar images using fuzzy clustering and EM clustering[J]. IEEE Transactions on Geoscience and Remote Sensing, 2005, 43(3): 519–527. DOI: 10.1109/TGRS.2004.842108.
[26]	Abramowitz M and Stegun I A. Handbook of Mathmatical Functions[M]. New York: Dover Pub. Inc., 1968.
[27]	Conradsen K, Nielsen A A, Schou J, et al. A test statistic in the complex Wishart distribution and its application to change detection in polarimetric SAR data[J]. IEEE Transactions on Geoscience and Remote Sensing, 2003, 41(1): 4–19. DOI: 10.1109/TGRS.2002.808066.
[28]	Qin X X, Zou H X, Zhou S L, et al. Simulation of spatially correlated PolSAR images using inverse transform method[J]. Journal of Applied Remote Sensing, 2015, 9(1): 095082. DOI: 10.1117/1.JRS.9.095082.
[29]	Arbelaez P, Maire M, Fowlkes C, et al. Contour detection and hierarchical image segmentation[J]. IEEE Transactions on Pattern Analysis and Machine Intelligence, 2010, 33(5): 898–916.