SVMMAP Modeling of SAR Imagery for Unsupervised Segmentation with Bootstrap Sampling

Abstract:

A spatially variant mixture multiscale autoregressive prediction (SVMMAP) model is present, which was applied to segmentation of SAR imagery. General process is as follow: at first, by Bootstrap sampling technique a small representative set of pixels is selected; then, expectation maximization (EM) algorithm and least square (LS) estimation were used to estimate the model, and minimum description length (MDL) rule was employed to choose classification number; at last, Bayes classifier was used to segment image. For a simulated image of size 256×256, a segmentation accuracy of 99.76% was achieved. Besides, quantitative assessment was also presented about segmentation quality of images.

• Periodical: Applied Mechanics and Materials (Volume 614)
• Pages: 393-396
• DOI: https://doi.org/10.4028/www.scientific.net/AMM.614.393
• Online since: September 2014
• Authors: Yan Wei Ju*, Yan Zhang
• Keywords: EM Algorithm, SAR Imagery, Segmentation

References

[1] Hang Yu, Xiangrong Zhang, Shuang Wang, Biao Hou: Context-Based Hierarchical Unequal Merging for SAR Image Segmentation, Geoscience and Remote Sensing, IEEE Transactions on. 51(2): 995-1009 (2013).

DOI: 10.1109/tgrs.2012.2203604

[2] Xiang, D., Tang, T., Hu, C., Li, Y., Su, Y.: A Kernel Clustering Algorithm With Fuzzy Factor: Application to SAR Image Segmentation, Geoscience and Remote Sensing Letters, IEEE Transactions on. 11(7): 1290-1294 (2014).

DOI: 10.1109/lgrs.2013.2292820

[3] WEN, X. B. and TIAN, Z.: Mixture multiscale autoregressive modelling of SAR imagery for segmentation, Electronics Letters, 39, p.1272–1274 (2003).

DOI: 10.1049/el:20030805

[4] MCLACHLAN, G. and PEEL, D.: Finite mixture models(New York, USA: JOHN WILEY &SONS, 2000).

[5] BLEKAS et al: A spatially constrained mixture model for image segmentation, IEEE Translations on Neural Networks, 16, pp.494-498(2005).

DOI: 10.1109/tnn.2004.841773

[6] Ju, Yanwei, Chu, Xiaobin, Xu, Ge: Generalized multiscale Rayleigh likelihood ratio for SAR imagery segmentation, Radar Conference, 2009 IET International (Guillin, China, 2009), pp.1-4.

DOI: 10.1049/cp.2009.0159

[7] JAIN et al: Bootstrapping techniques for error estimation, IEEE Translations on Pattern Anal. Mach. Intell., PAMI-9(5), p.628–633(1987).

DOI: 10.1109/tpami.1987.4767957

[8] ZRIBI, M.: Non-parametric and unsupervised Bayesian classification with Bootstrap sampling, Image and Vision Computing, 22, p.1–8(2004).

DOI: 10.1016/s0262-8856(03)00136-7

[9] PERNKOPF, F. and BOUCHAFFR, A, D.: Genetic-based EM algorithm for learning Gaussian mixture models, IEEE Translations on Pattern Analysis and Machine Intelligence, 27, p.1344–1348(2005).

DOI: 10.1109/tpami.2005.162

[10] CAVES et al, Quantitative comparison of the performance of SAR segmentation algorithms, IEEE Translations on Image Processing, 7, p.1534–1546(1998).

DOI: 10.1109/83.725361