光流法目前以來已經有許多不同應用，像是動作識別、機器人或車輛導航、先進駕駛輔助系統(ADAS)、物件偵測及監控系統等。雖然目前已經有很多光流法被提出，準確度也愈來愈好，但相較之下運算複雜度也較高，使得無法應用在即時處理系統上。因此本論文提出了一個低成本且適合於硬體實現的階層式區塊匹配光流法，將連續的兩張影像輸入，高品質且即時的估計光流向量。
本論文所提出的階層式區塊匹配光流法，此演算法利用區塊匹配法來估計初始光流，並利用局部平滑約束來修正運動向量，另外有幾個特點：1)採取兩層影像金字塔並且不同層級採取不同的搜尋初始位置。2)提出新的SAD評估方式，並且在硬體實現上資源使用率可更低。3)利用均值濾波器來平滑光流，降低錯誤估計光流的區域。
最後與其他主流實現在硬體上的演算法以及近期提出硬體架構比較，實驗結果證明本論文可以在當前複雜資料集提供更好的結果，以及更低的硬體資源成本，且工作時脈可達到200MHz(5ns)，並在640×480解析度的影像上達到每秒37幀(FPS)即時處理的效能。本篇論文利用AEE做為光流評估指標。

The optical flow has been used in various applications, such as motion recognition, robot or vehicle navigation, advanced driver assistance systems (ADAS), object detection and video surveillance. Recent studies of optical flow can achieve higher accuracy, but increasing computational complexity is difficult to be implemented in the real-time processing system. Therefore, this paper present a low-cost optical flow algorithm based on hierarchical block matching and more suitable for hardware implementation. Input is accepts two consecutive images and output is the high-quality and real-time estimation of the optical flow vector.
The optical flow algorithm based on hierarchical block matching proposed in this paper uses the block matching method to estimate the initial optical flow and uses local smoothing constraints to refine the motion vector. And this method has the following three characteristics: 1) Take two layers of image pyramid and different levels take different initial search positions. 2) Propose a new SAD evaluation methodology, and the resource utilization can be lower in hardware implementation. 3) Use The mean filter to smooth the optical flow and reduce the area of the optical flow that is incorrectly estimated.
Compared with other algorithms that are more commonly implemented on hardware and the recent proposed hardware architecture, this paper can provide better results in the current complex datasets, as well as lower hardware resource costs. The design works with the processing rate of 200MHz(5ns), which is fast enough to process 640x480 resolution at 37 frames per second (FPS) in real time. This paper uses AEE as indicator of optical flow evaluation.

[1] H. Wang, A. Kläser, C. Schmid, C.-L. Liu, “Dense trajectories and motion boundary descriptors for action recognition,” Int. J. Comput. Vision, vol. 103, no. 2, pp. 60–79, 2013.
[2] H. Chao, Y. Gu, M. Napolitano, “A survey of optical flow techniques for robotics navigation applications,” J. Intell. Robotic Syst., vol. 73, no. 1–4 , pp. 361–372, 2014.
[3] W. Enkelmann, “Obstacle detection by evaluation of optical flow fields from image sequences,” Image Vis. Comput., vol. 9, no. 3, pp. 160–168, 1991.
[4] Z. Sun, G. Bebis, R. Miller “On-road vehicle detection: a review,” IEEE Trans. Pattern Anal. Mach. Intell., vol. 28, no. 5, pp. 694-711, 2006.
[5] P. Sundberg, T. Brox, M. Maire, P. Arbeláez, and J. Malik, “Occlusion boundary detection and figure/ground assignment from optical flow,” in Proc. IEEE Conf. Comput. Vis. Pattern Recognit., Jun. 2011, pp. 2233–2240.
[6] A. Wali and A. M. Alimi, “Event detection from video surveillance data based on optical flow histogram and high-level feature extraction,” in Proc. 20th Int. Workshop Database Expert Syst. Appl., 2009, pp. 221–225.
[7] K. Pauwels and M. M. V. Hulle, “Realtime phase-based optical flow on the GPU,” in Proc. IEEE Comput. Soc. Conf. Comput. Vision Pattern Recog. Workshops (CVPRW), 2008, pp. 1–8.
[8] J. L. Barron, D. J. Fleet, and S. S. Beauchemin, “Performance of optical flow techniques,” Int. J. Comput. Vis., vol. 12, no. 1, pp. 43–77, 1994.
[9] B. D. Lucas and T. Kanade, “An iterative image registration technique with an application to stereo vision,” in Proc. IJCAI, vol. 2. 1981, pp. 674–679.
[10] B. K. Horn and B. G. Schunck, “Determining optical flow,” in Proc. Tech. Symp. EAST, 1981, pp. 319–331.
[11] H.-S. Seong, C. E. Rhee, and H.-J. Lee, “A novel hardware architecture of the Lucas–Kanade optical flow for reduced frame memory access,” IEEE Trans. Circuits Syst. Video Technol., vol. 26, no. 6, pp. 1187–1199, Jun. 2016.
[12] V. Mahalingam, K. Bhattacharya, N. Ranganathan, H. Chakravarthula, R. R. Murphy, and K. S. Pratt, “A VLSI architecture and algorithm for Lucas–Kanade-based optical flow computation,” IEEE Trans. Very Large Scale Integr. (VLSI) Syst., vol. 18, no. 1, pp. 29–38, Jan. 2010.
[13] F. Barranco, M. Tomasi, J. Díaz, M. Vanegas, and E. Ros, “Parallel architecture for hierarchical optical flow estimation based on FPGA,” IEEE Trans. Very Large Scale Integr. (VLSI) Syst., vol. 20, no. 6, pp. 1058–1067, Jun. 2012.
[14] S. Baker, D. Scharstein, J. P. Lewis, S. Roth, M. J. Black, and R. Szeliski, “A database and evaluation methodology for optical flow,” Int. J. Comput. Vis., vol. 92, no. 1, pp. 1–31, 2011.
[15] K. Seyid, A. Richaud, R. Capoccia and Y. Leblebici, "FPGA-Based Hardware Implementation of Real-Time Optical Flow Calculation," in IEEE Transactions on Circuits and Systems for Video Technology, vol. 28, no. 1, pp. 206-216, Jan. 2018.
[16] M. Bierling, “Displacement estimation by hierarchical blockmatching,” Proc. SPIE, vol. 1001, pp. 942–953, Oct. 1988.
[17] M. Santoro, G. Al-Regib, and Y. Altunbasak, “Smoothness Constraints in Recursive Search Motion Estimation for Picture Rate Conversion,” in Proc. VISAPP (2), 2012, pp. 399–403.
[18] C. Bartels and G. De Haan, “Smoothness constraints in recursive search motion estimation for picture rate conversion,” Circuits and Systems for Video Technology, IEEE Transactions on, vol. 20, no. 10, pp. 1310– 1319, 2010.
[19] T. Komarek and P. Pirsch, “Array architectures for block matching algorithms,” IEEE Trans. Circuits Syst., vol. 36, no. 10, pp. 1301–1308, Oct. 1989.
[20] D. D. Gajski, Principles of Digital Design. Upper Saddle River, NJ: Prentice-Hall, 1997.
[21] Middlebury Website. [Online].Available: http://vision.middlebury.edu/flow/data/
[22] P. Dollár, Piotr’s Computer Vision MATLAB Toolbox (PMT). [Online].Available: http://vision.ucsd.edu/~pdollar/toolbox/doc/index.html
[23] M. Tomasi, M. Vanegas, F. Barranco, J. Díaz, and E. Ros, “Highperformance optical-flow architecture based on a multi-scale, multiorientation phase-based model,” IEEE Trans. Circuits Syst. Video Technol., vol. 20, no. 12, pp. 1797–1807, Dec. 2010.
[24] M. Kunz, A. Ostrowski, and P. Zipf, “An FPGA-optimized architecture of Horn and Schunck optical flow algorithm for real-time applications,” in Proc. 24th Int. Conf. Field Program. Logic Appl. (FPL), Sep. 2014, pp. 1–4.