近年來，特徵影像描述子在各方面獲得了成功的應用，如：物件辨識 (Object recognition) 、紋理辨識 (Texture recognition)、影像檢索 (Image retrieval) 、機器人定位(Robot localization) 、影片資料探勘 (Video data mining) 等方面。但在即時性方面，卻因複雜度較高，以軟體實作方面較不容易達到即時性的要求。因此為了達到即時性的效果，我們以硬體實作的方式來加速。在本論文中將以FAST（Features from Accelerated Segment Test）和HOG (Histograms of Oriented Gradients)兩個特徵擷取演算法為研究目標，提出新的硬體設計方式來加快運算。
由於FAST最近正式被Khronos Group 採用作為OpenVX API，相關的硬體設計論文較少，因此在本研究中將提出一個新的實作方法，再分析硬體執行時間成本與硬體面積成本之間的關係，找出可達即時性下之最小硬體。而HOG計算需用到開根號和三角函數，因此所需的計算複雜度很高，在硬體設計方面較難以實現，在本論文中根據一些相關研究以近似的方式來取代複雜的運算，提出了能夠達到即時性且面積成本最小的硬體。

The description of local features of images has been successfully applied to many areas, including object recognition, texture recognition, image retrieval, robot localization, video data mining, etc. However, pure software implementations usually cannot achieve the requirement of real-time processing. In this thesis, we present hardware acceleration for two popular image feature extraction/description algorithms, Features from Accelerated Segment Test (FAST) and Histograms of Oriented Gradients (HOG). FAST is recently adopted as OpenVX API for image and vision processing. Since there are few hardware implementation papers, this thesis will propose a new design and analyzes the tradeoff between execution time, quality, and hardware cost, in order to find a suitable hardware acceleration. HOG computation requires complicated arithmetic operations such as square root and trigonometric function, making it costly for hardware implementation. In this thesis, we modify the original HOG algorithm in order to significantly reduce the hardware complexity without sacrificing too much in speed and quality.

論文審定書 i
公開授權書 ii
摘要 iii
英文摘要 iv
圖目錄（List of Figures） viii
表目錄（List of Tables） xi
第1章 緒論（Introduction） 1
1.1 研究動機 1
1.2 本文大鋼 2
第2章 研究背景與相關研究 3
2.1 文獻回顧 3
2.2 局部二值化Local Binary Pattern(LBP) 5
2.3 LBP與HOG結合 6
第3章 特徵擷取演算法 9
3.1 SIFT（scale-space extrema detection） 10
3.1.1 尺度空間極值偵測（scale-space extrema detection） 11
3.1.2 特徵點定位與篩選（keypoint localization） 13
3.1.4 特徵點描述（Keypoint descriptor） 16
3.1.5 應用(application) 17
3.2 SURF（Speeded Up Robust Features） 18
3.2.1 積分影像（Integral image） 20
3.2.2 快速 Hessian 特徵檢測（Fast Hessian detection） 21
3.2.3 特徵點方向性（Orientation assignment） 25
3.2.4 SURF特徵描述（Keypoint descriptor） 26
3.2.5 應用(application) 27
3.3 FAST (Features from Accelerated Segment Test) 28
3.3.1 加速方法 (Speed Up) 29
3.3.2 非最大值抑制 (Non-Maximal Suppression) 30
3.3.3 應用 (application) 31
3.4 HOG (Histograms of Oriented Gradients) 32
3.4.1 梯度計算 (Gradient Computation) 33
3.4.2 梯度投票 (Gradient Vote) 34
3.4.3 正規化 (Normalization) 35
3.4.4 應用 (application) 37
第4章 FAST/HOG硬體加速 39
4.1 FAST硬體設計 39
4.1.1 RGB to Intensity 39
4.1.2 Intensity Image Line Buffer 40
4.1.3 FAST Operator 41
4.1.4 Non-maximal Suppression Line Buffer 44
4.1.5 Non-maximal Suppression Operator 45
4.2 HOG硬體設計(第一版) 46
4.2.1 RGB to Intensity 46
4.2.2 Intensity Image Line Buffer 47
4.2.3 Gradient Computation 47
4.2.4 Gradient Vote 51
4.2.5 Normalization 53
4.3 HOG硬體設計(第二版) 53
4.3.1 Gradient Local Binary Pattern Computation 54
4.3.2 Binarization 56
第5章 實驗數據分析與比較 58
5.1 邏輯數據合成分析 58
5.2 比較(FAST) 60
5.3 比較(HOG、LBP-HOG、LBP-HOG+HOG、HOG V1V2) 61
第6章 結論與未來展望 65
6.1 結論 65
6.2 未來展望 65
第7章 參考文獻（References） 66

[1] D. G. Lowe, et al “Distinctive image features from scale-invariant keypoints,” Int. J. Computer Vision, vol. 60, no. 2, pp. 91-110, Jan. 2004.
[2] H. Bay, A. Ess, T. Tuytelaars, and L. Van Gool, “SURF: speeded-up robust features,” Comput. Vis. Image Underst., vol. 110, no. 3, pp. 346–359, Jun. 2008.
[3] P.-Y. Chen, et al., “An Efficient Hardware Implementation of HOG Feature Extraction for Human Detection,” IEEETrans. On Intelligent Transportation Systems, vol. 15, pp. 656-662, Apr. 2014.
[4] S-Y Lee, et al, “HOG feature extractor circuit for real-time human and vehicle detection,” TENCON 2012 - 2012 IEEE Region 10 Conference, 2012.
[5] T.H Dinh, et al, “High throughput FPGA architecture for corner detection in traffic images,” 2014 IEEE Fifth International Conference on ICCE, pp. 297–302, 2014.
[6] S-F Hsiao, and P-S Wu, “Acceleration of Image Feature Extraction Algorithms, Department of Computer Science of National Sun Yat-Sen University Master Thesis, 2014.
[7] J.J. Koenderink, “The structure of images,” Biological Cybernetics, vol. 50, no. 5, pp. 363-396, 1984.
[8] T. Lindeberg, “Scale-space theory: A basic tool for analysing structures at different scales,” Applied Statistics, vol. 21, no. 2, pp. 224-270, 1994.
[9] K. Mikolajczyk and C. Schmid. “A performance evaluation of local descriptors,” in Proc. Computer Vision and Pattern Recognition, Jun. 2003.
[10] K. Mikolajczyk and C. Schmid, “A performance evaluation of local descriptors,” IEEE Trans. Pattern Anal. Mach. Intell., vol. 27, no. 10, pp. 1615–1630, Oct. 2005
[11] M. Schaeferling, G. Kiefer, "Object Recognition on a Chip: A Complete SURF-based System on a Single FPGA". International Conference on Reconfigurable Computing and FPGAs, Nov. 2011, pp.49-54.
[12] J. Švab, T. Krajnik, J. Faigl, L. Preucil, "FPGA based Speeded up Robust Features". IEEE International Conference on TePRA 2009, pp.35-41.
[13] T. Sledevi_, A. Serackis,” SURF Algorithm Implementation on FPGA”. 13th Biennial Baltic Electronics Conference (BEC2012), Oct. 2012
[14] Rosten, et al” Faster and Better: A Machine Learning Approach to Corner Detection”. Pattern Analysis and Machine Intelligence, IEEE Transactions on , vol. 32, pp. 105-119, 2010.
[15] M. Schaeferling, G. Kiefer, "Object Recognition on a Chip: A Complete SURF-based System on a Single FPGA". International Conference on Reconfigurable Computing and FPGAs, Nov. 2011, pp.49-54.
[16] M. Brown and D. G. Lowe, “Invariant features from interest point groups,” in British Machine Vision Conf., pp. 656-665, 2002.
[17] E. Rosten and T. Drummond, “Machine Learning for High Speed Corner Detection,” Proc. Ninth European Conf. Computer Vision, vol. 1, pp. 430-443, 2006.
[18] N. Dalal and B. Triggs, “Histograms of Oriented Gradients for Human Detection,” Proc. IEEE Conf. Computer Vision and Pattern Recognition, vol. 1, pp. 886-893, July 2005.
[19] R. Kadota, H. Sugano, M. Hiromoto, H. Ochi, R. Miyamoto, and Y. Nakamura, “Hardware architecture for HOG feature extraction,” in Proc. IEEE Conf. Intell. Inf. Hiding Multimedia Signal Process, Nov. 2009, pp. 1330–1333.
[20] M. Hemmati, et al., “HOG Feature Extractor Hardware Accelerator for Real-time Pedestrian Detection,” Proc. Euromicro Conference on Digital System Design (DSD), pp. 543-550, Aug. 2014.
[21] S. Lee, et al., “HOG Feature Extractor Circuit for Real-time Human and Vehicle Detection,” Proc. TENCON 2012 - 2012 IEEE Region 10 Conference, pp.1-5, Nov. 2012.
[22] DC. He and L. Wang , "Texture Unit, Texture Spectrum, And Texture Analysis", Proc. IEEE Transactions on Geoscience and Remote Sensing , vol. 28, pp. 509 – 512,1990.
[23] Ning Jiang et al , A. Serackis, “Gradient Local Binary Patterns for human detection,” Proc. IEEE International Symposium on Circuits and Systems (ISCAS), pp. 978-981, May 2013.
[24] K. Mizuno, Y. Terachi, K. Takagi, S. Izumi, H. Kawaguchi, and M.Yoshimoto, "Architectural Study of HOG Feature Extraction Processor for Real-Time Object Detection," in Proc. IEEE Workshop on Signal Processing Systems (SiPS), 2012, pp. 197-202.
[25] M. Komorkiewicz, M. Kluczewski, and M. Gorgon, "Floating point HOG implementation for real-time multiple object detection," in Proc. 22nd Int. Conf. Field Programmable Logic and Applications (FPL), 2012, pp. 711-714.
[26] K. Negi, K. Dohi, Y. Shibata, and K. Oguri, "Deep pipelined one-chip FPGA implementation of a real-time image-based human detection algorithm," in Proc. Int. Conf. Field-Programmable Technology (FPT), 2011, pp. 1-8.
[27] Seongwook Park; Gyeonghoon Kim; Junyoung Park; Hoi-Jun Yoo, " A 1.5nJ/pixel super-resolution enhanced FAST corner detection processor for high accuracy AR" ESSCIRC 2014 - 40th.
[28] Dohi, K.; Yorita, Y.; Shibata, Y.; Oguri, K., " Pattern Compression of FAST Corner Detection for Efficient Hardware Implementation" International Conference on Field Programmable Logic and Applications (FPL), 2011.
[29] Marek Kraft, Adam Schmidt and Andrzej Kasi´nski, " High-speed image feature detection using fpga implementation of fast algorithm" International Conference on Computer Vision Theory and Applications (VISAPP), 2008.
[30] INRIA person dataset. [Online]. Available: http://pascal.inrialpes.fr/data/.
[31] C. C. Chang, and C. J. Lin, “LIBSVM -- A Library for Support Vector Machines, ” http://www.csie.ntu.edu.tw/~cjlin/libsvm/
[32] T. Ojala, M. Pietikäinen, and D. Harwood, "Performance evaluation of texture measures with classification based on Kullback discrimination of distributions", Proceedings of the 12th IAPR International Conference on Pattern Recognition (ICPR 1994), vol. 1, pp. 582 - 585.
[33] Xiaoyu Wang; Han, T.X. and Shuicheng Yan, "An HOG-LBP human detector with partial occlusion handling", IEEE 12th International Conference on Computer Vision, 2009.
[34] Junqiang Wang; Huadong Ma, "MPL-Boosted Integrable Features Pool for pedestrian detection", IEEE 18th International Conference on Image Processing (ICIP), 2011.
[35] Junqiang Wang; Huadong Ma and Anlong Ming, " Fast accurate pedestrian detection using a MPL-Boosted cascade of weak FIK-SVM classifiers", IEEE International Conference on Multimedia and Expo (ICME), 2011.
[36] Chengbin Zeng; Huadong Ma and Anlong Ming, " Fast human detection using mi-sVM and a cascade of HOG-LBP features", IEEE International Conference on Image Processing (ICIP), 2010.
[37] Chengbin Zeng and Huadong Ma, " Robust Head-Shoulder Detection by PCA-Based Multilevel HOG-LBP Detector for People Counting", IEEE International Conference on Pattern Recognition (ICPR), 2010.
[38] Guolong Gan and Jian Cheng, " Pedestrian Detection Based on HOG-LBP Feature", IEEE International Conference on Computational Intelligence and Security (CIS), 2011.
[39] Qing, Chunmei; Dickinson, P.; Lawson, Shaun and Freeman, Robin, " Automatic nesting seabird detection based on boosted HOG-LBP descriptors", IEEE International Conference on Image Processing (ICIP), 2011.
[40] Long Ma; Chunheng Wang and Baihua Xiao, " Text detection in natural images based on multi-scale edge detetion and classification ", International Congress on Image and Signal Processing (CISP), 2010.