臺灣博碩士論文加值系統

本論文主要是針對影像處理上兩個重要的議題加以探討，首先是影像的比對，其目的是在目標影像中辨識出參考影像，在亮度和顏色各種差異情況下；第二是影像的還原，試著使用DSP技術的方法去復原模糊影像，因為錯誤的對焦或是手震所造成的模糊。
在影像比對部分，採用了標準化互相關(NCC)作比對方式，為了完成即時操作，NCC運算方法實現在Xilinx Virtex-6 XC6VLX240T的FPGA，並且使用內建FPGA嵌入式的乘法器DSP48E1s，減少FPGA slice使用資源以及增加運算速度。為了實現較低的複雜度，簡化NCC演算法來避免成本較高的模組，例如開根號計算。為了達到100MHz工作頻率的目標，使用管線化架構之處理和各式的電路優化技巧，最後完成的硬體電路的工作頻率為154.405MHz，吞吐量為19.1Mpixel/s。
在影像還原部分，我們主要針對失焦或手震兩種情形造成的模糊影像，假設模糊影像為高斯模型，還原方法開始先分析模糊影像的頻譜特性來估測模糊影像模型的模糊參數，應用快速傅立葉轉換獲得頻譜影像，找出頻譜特性跟模糊參數之間的聯繫關係。為了更容易估測參數，將頻譜影像進一步做積分和設定門檻值做量化，量化成一個簡單的二值化影像，藉由分析二值化影像圖形的走向趨勢，高斯分佈的PSF參數可以更容易獲得。為了簡化估測參數的過程，設計出一個查詢表格，用來轉換積分二值化頻譜參數所對應到的模糊參數，之後使用Wiener Filter做還原影像，不用考慮雜訊被增強的問題。此提出的方法已可以成功應用在各式影像，如風景照、近物照和人物照，實驗的還原影像結果可以有效的改善PSNR值。

This thesis focuses on two important topics in image processing. The first one is the image matching, its purpose is to identify a target image from a reference image in the presence of various discrepancy factors such as brightness and color. The second is the image restoration, which tries to recover blurred images due to improper focus or hand movement by way of DSP techniques.
In image matching, a normalized cross correlation (NCC) index is adopted as the matching factor. To achieve real time operations, the NCC computing scheme is hard-wired implemented in a Xilinx Virtex-6 XC6VLX240T FPGA. The built-in FPGA embedded multipliers DSP48E1s are used to reduce the usage of FPGA slices and to increase the processing speed. To lower the implementation complexity, the NCC algorithm is simplified to avoid hardware expensive modules such as the square root computation. To meet the target 100MHz working frequency, extensive pipelining measures and various circuit optimization techniques are employed. Final implementation result indicates the design is capable of working up to 154.405MHz, and the equivalent throughput is 19.1 Mpixel / s.
In image restoration, we focus on the blurred images caused by defocusing or hand movement. A Gaussian blurred image model is assumed. The restoration scheme starts with analyzing the spectral characteristic of a blurred image to estimate the parameters of the blurred image model. The spectral image is first obtained by applying a fast Fourier transform and the linkage between the spectral characteristics and the parameters are identified. To ease the parameter estimation process, the spectral image is further integrated and quantized by a threshold value, which leads to a simple binary image. By analyzing the pattern and the orientation of the binary image, the parameters of the Gaussian distribution based point spread function can be easily obtained. To simplify the estimation process, a lookup table is devised, which translates the spectral parameters of binary integral image to the corresponding blurred function parameters. After that, a Wiener Filter is employed to restore the image without introducing the noise enhancement problem. The proposed scheme has been successfully applied to various images classified as landscapes, near objects and portraits. The experimental results show significant PSNR improvements in the restored images.

摘要 i
Abstract ii
目錄 iii
圖目錄 v
表目錄 vii
第一章 緒論 1
1.1 研究背景 1
1.2 研究目的及動機 1
1.3 論文架構 2
第二章 影像比對及成像系統背景 3
2.1 影像比對介紹 3
2.1.1 尺度不變特徵轉換 3
2.1.2 標準互相關 5
2.2 影像成像介紹 6
2.2.1 成像原理 6
2.2.2 焦點及景深概念 6
2.2.3 影像退化系統基本定義 7
2.2.4 影像退化系統的矩陣描述 8
2.2.5 點擴散函數 9
第三章 高速率與彈性影像比對系統設計與實現 11
3.1 動作概述 11
3.2 演算法化簡及模擬結果 12
3.3 FPGA的硬體加速設計 18
3.3.1 規格要求及分析 18
3.3.2 演算法映射硬體 19
3.3.3 FPGA內建資源分析加速 25
3.4 FPGA執行結果及系統平台 26
第四章 模糊影像還原回顧 28
4.1 還原演算法概述 28
4.2 代數還原法 28
4.2.1 無條件限制性還原 29
4.2.2 有條件限制性還原 29
4.3 反濾波法(Inverse Filter) 30
4.4 最小平方濾波器(Least Squares Filter) 31
4.5 限制性最小平方濾波器(Constrained Least Squares) 33
4.6 截斷限制性最小平方濾波器(Truncated Constrained Least Squares) 37
4.7 盲目影像還原技術 40
4.7.1 ARMA參數估測法 41
4.7.2 疊代盲目去迴旋積法 43
4.8 影像還原演算法小結 44
第五章 新穎模糊影像還原法及效能評估 45
5.1 前言 45
5.2 新穎模糊影像還原演算法 45
5.2.1 影像前處理 46
5.2.2 模糊參數辨認 50
5.2.3 模糊影像還原 52
5.3 新穎模糊影像還原演算法模擬測試 53
第六章 結論也未來展望 63
參考文獻 64

[1]Y. C. Trivedi and L. Kurz, “Image restoration using recursive estimators,” IEEE Trans. Syst., Man., Cybern., vol. 25, pp. 1470–1482, Nov.1995.
[2]M. , Chellali and V.K. , Ingle“ Adaptive image restoration using the multichannel recursive least squares algorithm,” Conference on Acoustics, Speech, and Signal Processing (ICASSP), Apr 1991 pp.14-17.
[3]蔡憲裕，“半自動化異元高解析力遙測影像加強適應性套和研究”，國立中央大學碩士論文，民國九十九年。
[4]Jiun-Yan Chen, Kuo-Feng Hung, Hsin-Yi Lin, Yen-Chung Chang, Yin-Tsung Hwang, Ciao-Kai Yu, Cheng-Ru Hong, Chin-Chia Wu, Yung-Jung Chang, “Real-time FPGA-based Template Matching Module for Visual Inspection Aplication”, IEEE/ASME International Conference on Advanced Intelligent Mechatronics, , July 2012 pp.11-14.
[5]Mark S. Nixon and Alberto S. Aguado. “Feature Extraction and Image Processing.” Academic Press, 2008, pp. 88.
[6]Lihong Yang and Jianyue Ren, “Remote Sensing Image Restoration using Estimated Point Spread Function,” 2010 International Conference on Information, Networking and Automation, 2011.
[7]Tao Kong, Chunguang Xu, Dingguo Xiao, “Experimental ESD method for restoration of blurry image in ultrasonic C-Scan,” 2010 International Conference on Mechanic Automation and Control Engineering (MACE), June 2010, pp. 26-28.
[8]Feng-qing Qin, Jun Min, Hong-rong Guo, “A Blind Image Restoration Method Based on PSF Estimation.”, World Congress on Software Engineering, 2009 IEEE.
[9]D. SRINIVASA RAO, K. SELVANI DEEPTHI, K. MONI SUSHMA DEEP, “Application of Blind Deconvolution Algorithm for Image Restoration,” International Journal of Engineering Science and Technology, Jan. 2011.
[10]Rob Fergus, Barun Singh, Aaron Hertzmann, Sam T. Roweis, William T. Freeman, “Removing Camera Shake from a Single Photograph,” [Online]
Available:http://cs.nyu.edu/~fergus/papers/deblur_fergus.pdf
[11]Feng Duan, Tanning Zhang, “The Estimation of Blur Based on Image Information,” Fifth International Conference on Image and Graphics, 2009. ICIG ''09. Sept. 2009, pp.109-112.
[12]Sun Shaojie, Wu Qiong, Li Guohui, “Image restoration for single blurred image,” IEEE International Conference on Intelligent Computing and Intelligent Systems, 2009. ICIS 2009. Nov. 2009, vol.4 pp.491-494.
[13]Fredenc Guichard, Hoang Phi Nguyen, Regis Tessieres, Marine Pyanet, Imene Tarchouna, Frederic Cao, “Extended depth-of-field using sharpness transport across color channels.”[Online]
Available:http://www.dxo.com/sites/dump.dxo.com/files/dxoimages/ei/sci-publications/2009_EI_EDOF.pdf
[14]O. Cossairt and S. Nayar, “Spectral focal sweep: Extended depth of field from chromatic aberrations,” in Proc. Int. Conf. Comput. Photography, 2010, pp. 1–8.
[15]O. Cossairt, C. Zhou, and S. K. Nayar, “Diffusion Coding Photography for Extended Depth of Field,” ACM Trans. on Graphics, Aug. 2010.
[16]蕭進松，“數位影像處理”，全華科技圖書股份有限公司。
[17]繆紹綱，“數位影像處理-活用Matlab”， 全華科技圖書股份有限公司。
[18]S. Kim, S. Jun, E. Lee, J. Shin, and J. Paik, “Real-time bayer-domain image restoration for an extended depth of field (EDof) camera,” IEEE Trans. Consumer Electron., vol. 55, no. 4, pp. 1756-1764, Nov. 2009.
[19]Jinyoung Youn, Younguk Park, Jeongho Shin, Joonki Paik, “Spatially Adaptive Image Restoration and Its FIR Implementation,” Second International Conference on Future Generation Communication and Networking Symposia, 2008. FGCNS ''08. Aug. 2010.
[20]T,A,Cheenama, I.M,Qureshi, and A.Hussain, “Blind Image Deconvolution Using Space-variant Neural Network Approach”, Electronics Letters, IEE-INST ELEC ENG, England, Mar. 2005, vol.41, pp.308-309.
[21]Canny, John, “A Computational Approach to Edge Detection,” IEEE Transactions on Pattern Analysis and Machine Intelligence, Nov. 1986, pp. 679-698.