臺灣博碩士論文加值系統

立體視覺(Stereo Vision)是一種從平面圖像中恢復深度信息的技術，僅需使用兩台相機在同一水平線上個別擷取場景，就能藉由類似人類雙眼的原理估計每個像素的視差，再轉成深度資訊，具有實現簡單，成本低廉，並且可以在非接觸條件下測量距離等優點。目前生活中已有許多地方應用此技術，像是機器人的導航判斷、目標拾取，自動工業化的部件安裝，自動駕駛距離偵測等等。

在演算法方面分為區域性和全域性，一般來說區域性演算法有速度快、計算量小的優點，而全域性演算法則可以得到更準確的深度資訊。為了減少成本以及達到即時的效果，我們選擇區域性演算法。立體匹配演算法分為四個階段：匹配代價計算(Matching Cost Computation)、代價聚合(Cost Aggregation)、視差選擇(Disparity Selection)、視差優化(Disparity Refinement)，因匹配代價階段得到的是圖像上每一個區塊的代價，但是這些值都是孤立的，引入過多的雜訊，本文使用引導影像濾波器(Guided Image Filter)在代價聚合階段去除雜訊，且我們拿掉了一半的引導影像濾波單元，減少一半的視差計算，並在視差優化階段更新算法以降低結果錯誤率，此架構在Xilinx Virtex-7 FPGA上實作，並且使用Middlebury上的四張標準圖(Tsukuba,Venus,Teddy,Cones)進行評估，我們提出的方法跟現有設計相比可以在少量影響準確率的情況下得到更少的硬體成本使用量。

Stereo Vision is a technique for recovering depth information from a planar image. By using two cameras to capture scenes on the same horizontal line, it is possible to estimate the disparity of each pixel by the principle similar to human eyes. And turn into depth information, which has the advantages of simple implementation, low cost, and the ability to measure distance under non-contact conditions. There are many places in the life that apply this technology, such as robot navigation judgement, target picking, automatic industrial component installation, automatic driving distance detection and so on.

Stereo matching algorithms can be classified into two categories: local and global. Local algorithms have the advantages of fast speed and small amount of calculation, while global algorithms can obtain more accurate depth information. To reduce costs and achieve real-time results, we choose local algorithms. Stereo matching algorithm is divided into four stages: Matching Cost Computation, Cost Aggregation, Disparity Selection, and Disparity Refinement. The matching cost phase gets the cost of each block. These values are isolated and have too much noise. Therefore, the Guided Image Filter is used to remove noise during the cost aggregation phase. And we reduced the half number of guided image filtering units and disparity calculations, and updated the algorithm to reduce the resulting error rate during the disparity refinement phase. This architecture was implemented on a Xilinx Virtex-7 FPGA and was evaluated against four standard Middlebury Benchmark image pairs viz.Tsukuba, Venus, Teddy, and Cones. The proposed methodology can get less hardware cost with similar accuracy which compared with state-of-the-art designs.

摘要 II
Abstract III
Content IV
Figure Captions VI
Table Captions VIII
Chapter 1. Introduction 1
1.1 Background 1
1.2 Motivation 2
1.3 Paper structure 3
Chapter 2. Related Work 4
2.1 Stereo matching principle 4
2.2 Stereo matching algorithm 5
2.2.1 Local matching algorithm 5
2.2.2 Global matching algorithm 7
2.3 Stereo matching step 7
2.3.1 Matching Cost Computation 8
2.3.2 Cost Aggregation 8
2.3.3 Disparity Selection 9
2.3.4 Disparity Refinement 9
Chapter 3. Proposed Algorithm 11
3.1 High-performance low-cost disparity prediction algorithm based on region similarity 11
3.2 Algorithm step 13
3.2.1 Matching Cost Computation 13
3.2.2 Cost Aggregation 15
3.2.3 Disparity Selection 16
3.2.4 Disparity Refinement 17
Chapter 4. Hardware Architecture Design 20
4.1 Mean filtering 20
4.2 GCMMU 21
4.2.1 GCC 21
4.3 CVCFDSU 22
4.3.1 CCU 22
4.3.2 Mean filter hardware 23
4.3.3 GIF 24
4.3.4 WTA 27
4.4 DRU 27
4.4.1 LR Check ＆ Filling 27
4.4.2 MedianY Filtering 28
Chapter 5. Experimental Results 30
5.1 Test data 30
5.2 Experimental data and results 31
Chapter 6. Conclusion 35
6.1 Conclusion 35
References 36

[1] D. Szeliski and R. Scharstein, A taxonomy and evaluation of dense two-frame stereo correspondence algorithms, Inter. J. of Comput. Vision, vol. 47, pp. 7-42, 2002.

[2] L. Nalpantidis, G. C. Sirakoulis, and A. Gasteratos, Review of stereo vision algorithms: from software to hardware, Inter. J. of Optomechatronics, vol. 2, no. 4, pp. 435-462, 2008.

[3] L. F. S. Cambuim, J. P. F. Barbosa, and E. N. S. Barros, “Hardware module for low-resource and real-time stereo vision engine using semi-global matching approach,“ in Proc. SBCCI, pp. 53–58, Aug./Sep. 2017.

[4] C. Banz, S. Hesselbarth, H. Flatt, H. Blume, and P. Pirsch, Real-time stereo vision system using semi-global matching disparity estimation: Architecture and fpga-implementation, in ICSAMOS, 2010, pp. 93-101.

[5] K.-J. Yoon and I.-S. Kweon, “Adaptive Support-Weight Approach for Correspondence Search,IEEE Trans. Pattern Matching and Machine Intelligence,vol. 28, no. 4, pp. 650-656, Apr. 2006.

[6] F. Tombari, S. Mattoccia, and L. D. Stefano, “Segmentation based adaptive support for accurate stereo correspondence, in Proc. Pacific-Rim Symp. Image Video Technol., pp. 427–438, Jun. 2006.

[7] N. Chang, T. Tsai, B. Hsu, Y. Chen, and T. Chang, Algorithm and Architecture of Disparity Estimation With Mini-Census Adaptive Support Weight, IEEE Trans. on CSVT, vol. 20, no. 6, pp. 792-805, 2010.

[8] J. Ding et al., Real-time stereo vision system using adaptive weight cost aggregation approach, EURASIP JIVP, 2011.

[9] S. Perri, P. Corsonello, and G. Cocorullo, Adaptive Census Transform: A novel hardware-oriented stereovision algorithm, Computer Vision and Image Understanding, vol. 117, no. 1, pp. 29-41, January 2013.

[10] Kaiming He, Jian Sun, and Xiaoou Tang, Guided Image Filtering, IEEE Trans. Pattern Anal. Mach. Intell., vol. 35, no. 6, pp. 1397-1409, 2013.

[11] A. Hosni, M. Bleyer, C. Rhemann, M. Gelautz, and C. Rother,“Real-time local stereo matching using guided image filtering, inProc. IEEE Int. Conf. Multimedia Expo, pp. 1–6, 2011.

[12] C. Ttofis and T. Theocharides, “High-quality real-time hardware stereo matching based on guided image filtering, inProc. Des.,Autom. Test Eur. Conf. Exhib., pp. 1–6, 24-28 Mar. 2014.

[13] C. Ttofis, C. Kyrkou, and T. Theocharides, “A low-cost real-timeembedded stereo vision system for accurate disparity estimation basedon guided image filtering,IEEE Trans. Comput., vol. 65, no. 9,pp. 2678–2693, Sep. 2016.
[14] C.K. Vala, K. Immadisetty, A. Acharyya, C. Leech, V. Balagopal, G.V. Merrett, ＆ B.M. Al-Hashimi, “High-speed low-complexity guided image filtering-based disparity estimation, IEEE Transactions on Circuits and Systems I: Regular Papers, vol. 65, Issue 2, pp.606–617, 2018.

[15] B. Salehian and A. M. Fotouhi, “Dynamic programming-based dense stereo matching improvement using an efficient search space reduction technique, in Optik, vol. 160, pp.1-12, 2018

[16] J. Sun,H.Y. Shum,and N.N.Zheng, “Stereo matching using belief propagation, IEEE Transactions on Pattern Analysis and Machine Intelligence, 25(7):787-800,July 2003.

[17] P. F. Felzenszwalb and D.P. Huttenlocher, “Efcient belief propagation for early vision, In IEEE Conference on Computer Vision and Pattern Recognition, 2004.

[18] T. Taniai, Y. Matsushita, and T. Naemura, “Graph Cut Based Continuous Stereo Matching Using Locally Shared Labels, In IEEE Conference on Computer Vision and Pattern Recog-nition, pages 1613–1620, 2014.
[19] X. Huang, C. Yuan, and J. Zhang, “Graph cuts stereo matching based on patch-match and ground control points constraint, In Advances in Multimedia Information Processing–PCM 2015, pages 14–23. Springer,2015.

[20] K.-J. Yoon and I. S. Kweon, “Adaptive support-weight approach for correspondence search, IEEE Transactions on Pattern Analysis and Machine Intelligence, vol. 28, no. 4, pp. 650–656, 2006.
[21] Kaiming He, Jian Sun, and Xiaoou Tang, “Guided Image Filtering, IEEE Trans. Pattern Anal. Mach. Intell., vol. 35, no. 6, pp. 1397-1409, 2013.

[22] S. A. Fahmy, P. Y. K. Cheung, and W. Luk, “Novel fpga-based implementation of median and weighted median filtersfor image processing. In Proc. of the 2005 Int. Conf. on Field Programmable Logic and Applications (FPL). IEEE,pp. 142–147, 2005.

[23] L. A. Aranda, P. Reviriego, and J. A. MaestroError, “detection technique for a median filter IEEE Trans. Nuclear Sci., vol. 64, no. 8, pp. 2219–2226,Aug. 2017.

[24] J. Ding, J. Liu, W. Zhou, H. Yu, Y. Wang, and X. Gong, “Real-timestereo vision system using adaptive weight cost aggregation approach,EURASIP J. Image Video Process., vol. 2011, no. 1, p. 20, 2011.

[25] W. Wang, J. Yan, N. Xu, Y. Wang, and F.-H. Hsu, “Real-time high-quality stereo vision system in FPGA,IEEE Trans. Circuits Syst. VideoTechnol., vol. 25, no. 10, pp. 1696–1708, Oct. 2015.

[26] M. Jin and T. Maruyama, “Fast and accurate stereo vision system on FPGA,ACM Trans. Reconfigurable Technol. Syst., vol. 7, no. 1, p. 3,2014.

[27] Y. Shan et al., “Hardware acceleration for an accurate stereo vision sys-tem using mini-census adaptive support region,ACM Trans. EmbeddedComput. Syst., vol. 13, no. 4s, 2014, Art. no. 132.

[28] L. Zhang, K. Zhang, T. S. Chang, G. Lafruit, G. K. Kuzmanov, andD. Verkest, “Real-time high-definition stereo matching on FPGA, inProc. 19th ACM/SIGDA Int. Symp. Field Program. Gate Arrays, 2011,pp. 55–64.

[29] S. Jin et al., “FPGA design and implementation of a real-time stereo vision system,IEEE Trans. Circuits Syst. Video Technol., vol. 20, no. 1,pp. 15–26, Jan. 2010.