背景擷取對於電腦視覺和擴增實境是很重要的應用，現有多數的背景擷取法不適用於複雜前景移動的影片。本論文提出了一種基於混合高斯條件隨機場的背景擷取法，評估出最佳frame 的權重從固定視頻影片擷取出背景。此方法藉由intra-frame 和inter-frame 的關係，考慮像素間空間和時間相干性以及對比，作為frame 的權重。因為背景被假設為靜止的，本論文也提出一種基於時間相干性的無動量(motion-less) patch 擷取。此外，此論文也建議了一種客觀的評估性能法。實驗結果表示，此論文提出的方法比多數最先進的演算法更加有效率和穩健性。

Background extraction is important for applications in computer vision and augmented reality. Most existing methods are not suitable for video sequences containing complex foreground movement. Therefore, this work introduces a novel
extraction method based on joint Gaussian conditional random fields (JGCRF) to
estimate optimal frame weights for compositing a clear background from a fixed-view video sequence. The proposed algorithm analyzes the intra-frame and inter-frame relationship to consider spatial and temporal coherence and contrast distinctness among pixels for the basis of frame weights. Since a background is assumed to be static, a motion-less patch extractor is developed based on temporal coherence. Furthermore, an objective methodology is suggested in this work for performance evaluation. Quantitative and qualitative experimental results demonstrate the effectiveness and robustness of our approach over several state-of-art algorithms.

[1] L. Yang, H. Cheng, J. Su, and X. Li, “Pixel-to-model distance for robust background reconstruction,” IEEE Transactions on Circuits and Systems for Video
Technology, April 2015.
[2] A. Colombari and A. Fusiello, “Patch-based background initialization in heavily cluttered video,” IEEE Transactions on Image Processing, vol. 19, pp. 926–933, Apr. 2010.
[3] C.-C. Chen and J. K. Aggarwal, “An adaptive background model initialization
algorithm with objects moving at different depths,” IEEE International Conference on Image Processing, 2008.
[4] C. Stauffer and W. Grimson, “Adaptive background mixture models for realtime tracking,” in Proc. Conf. Comput. Vis. Pattern Recognit., vol. 2, pp. 246–252, Jun. 1999.
[5] Background Initialization website. [Online]. Available: http://www.diegm.
uniud.it/fusiello/demo/bkg/.
[6] Scene Background Initialization (SBI) dataset. [Online]. Available: http://
sbmi2015.na.icar.cnr.it/SBIdataset.html.
[7] J.-W. Seo and S. D. Kim, “Recursive on-line (2D)2PCA and its application to
long-term background subtraction,” IEEE Transactions on Multimedia, vol. 16,
pp. 2333–2344, Dec. 2014.
[8] J. Wen, Y. Xu, J. Tang, Y. Zhan, Z. Lai, and X. Guo, “Joint video frame
set division and low-rank decomposition for background subtraction,” IEEE Transactions on Circuits and Systems for Video Technology, vol. 24, pp. 2034–
2048, Dec. 2014.
[9] A. Staglianò, N. Noceti, A. Verri, and F. Odone, “Online space-variant background modeling with sparse coding,” IEEE Transactions on Image Processing,
vol. 24, pp. 2415–2428, Aug. 2015.
[10] X. Liu, G. Zhao, J. Yao, and C. Qi, “Background subtraction based on low-rank and structured sparse decomposition,” IEEE Transactions on Image Processing, vol. 24, pp. 2502–2514, Aug 2015.
[11] D. K. Panda and S. Meher, “Detection of moving objects using fuzzy color
difference histogram based background subtraction,” IEEE Signal Processing
Letters, vol. 23, pp. 45–49, Jan. 2016.
[12] X. Zhang, C. Zhu, S. Wang, Y. Liu, and M. Ye, “A bayesian approach for camouflaged moving object detection,” IEEE Transactions on Circuits and Systems
for Video Technology, April 2016.
[13] Scene Background Modeling and Initialization. [Online]. Available: http://
changedetection.net/.
[14] L. Maddalena and A. Petrosino, Background Modeling and Foreground Detection for Video Surveillance, ch. Background Model Initialization for Static Cameras, pp. 1–16. Chapman and Hall/CRC, 2014.
[15] K. Toyama, J. Krumm, B. Brumitt, and B. Meyers, “Wallflower: Principles
and practice of background maintenance,” IEEE International Conference on
Computer Vision, vol. 1, pp. 255–261, 1999.
[16] W. Long and Y. Yang, “Stationary background generation: an alternative to
the difference of two images,” Pattern Recognition, vol. 23, pp. 1351–1359, 1990.
[17] L. Li, W. Huang, I. Y.-H. Gu, and Q. Tian, “Statistical modeling of complex backgrounds for foreground object detection,” Image Processing, IEEE Transactions on, vol. 13, pp. 1459–1472, Nov. 2004.
[18] D. Russell and S. Gong, “A highly efficient block-based dynamic background
model,” in Proc. of IEEE Conference on Advanced Video and Signal Based
Surveillance, pp. 417–422, Sep. 2005.
[19] V. Reddy, C. Sanderson, and B. Lovell, “An efficient and robust sequential
algorithm for background estimation in video surveillance,” IEEE International
Conference on Image Processing, pp. 1109–1112, 2009.
[20] V. Reddy, C. Sanderson, and B. Lovell, “A low-complexity algorithm for static background estimation from cluttered image sequences in surveillance contexts,” Journal on Image and Video Processing, Apr. 2011.
[21] S.-S. Huang, L.-C. Fu, and P.-Y. Hsiao, “Region-level motion-based foreground segmentation under a bayesian network,” Circuits and Systems for Video Technology, IEEE Transactions on, vol. 19, pp. 522–532, Apr. 2009.
[22] L. Maddalena and A. Petrosino, “Towards benchmarking scene background initialization,” SBMI 2015 Workshop in conjunction with ICIAP 2015, September 2015.
[23] R. Shen, I. Cheng, and A. Basu, “QoE-based multi-exposure fusion in hierarchical multivariate Gaussian CRF,” Image Processing, IEEE Transactions on, vol. 22, no. 6, pp. 2469–2478, 2013.
[24] R. F. Tappen, C. Liu, E. H. Adelson, and W. T. Freema, “Learning Gaussian
conditional random fields for low-level vision,” IEEE Conference on Computer
Vision and Pattern Recognition, pp. 1–8, 2007.
[25] H. Rue and L. Held, Gaussian Markov Random Fields: Theory and Applications. Chapman and Hall/CRC, Boston, MA, 2005.
[26] P. Soille, Morphological Image Analysis: Principles and Applications. Springer-Verlag New York, Inc., 2nd ed., 2003.