(18.206.238.77) 您好!臺灣時間:2021/05/12 01:32
字體大小: 字級放大   字級縮小   預設字形  
回查詢結果

詳目顯示:::

我願授權國圖
: 
twitterline
研究生:魏世滿
研究生(外文):Osmand Prayitno
論文名稱:以SURF的方法為基礎使用多視角影像實現超解析度
論文名稱(外文):SURF-based Super Resolution using Multi-View Images
指導教授:繆紹綱繆紹綱引用關係
指導教授(外文):Shaou-Gang Miaou
學位類別:碩士
校院名稱:中原大學
系所名稱:電子工程研究所
學門:工程學門
學類:電資工程學類
論文種類:學術論文
論文出版年:2009
畢業學年度:97
語文別:英文
論文頁數:76
中文關鍵詞:影像對準SURF超解析影像重建
外文關鍵詞:image registrationSURFSuper resolutionimage reconstruction
相關次數:
  • 被引用被引用:0
  • 點閱點閱:296
  • 評分評分:系統版面圖檔系統版面圖檔系統版面圖檔系統版面圖檔系統版面圖檔
  • 下載下載:0
  • 收藏至我的研究室書目清單書目收藏:1
超解析影像是指從一組經過次像素位移的低解析度影像重建出高解析
度影像。每一個低解析度影像均包含有關景物的新資訊,而超解析的目的
就是要結合這些低解析度影像以得到更高解析度的影像。一般來說可分為
兩個主要工作:(1)影像對準是將拍攝同景物之兩個或更多的影像重疊的程
序;(2)影像重建是將低解析度影像重建成高解析度影像。本論文在影像對
準部份中使用SURF (Speeded Up Robust Features) 以改善高解析影像的品
質。接著本方法與Fourier Upsampling, SIFT (Scale-Invariant Feature
Transform), 和Keren 的方法比較。實驗結果顯示,使用SURF 可以比其
他方法執行的更好且更快。
Image super-resolution refers to the reconstruction of a high resolution
(HR) image from a set of low resolution (LR) images which are sub-pixel
shifted from each other. Each low resolution image contains new information
about the scene and super-resolution aims at combining these to give a higher
resolution image. In general, it can be broken into two major tasks: 1) image
registration which is the process of overlaying two or more images of the same
scene and 2) image reconstruction of the LR images into an HR image. This
thesis uses the SURF (Speeded Up Robust Features) for image registration to
improve the quality of the HR image. This method then compared with other
methods, named Fourier Upsampling, SIFT (Scale-Invariant Feature Transform),
and Keren. The experimental result shows that the SURF method can perform
better and faster than the other methods.
Abstract (in English) .............................................................................................. I
Abstract (in Chinese).............................................................................................II
Acknowledgements ............................................................................................. III
Table of Contents ................................................................................................ IV
List of Figures ....................................................................................................VII
List of Tables ....................................................................................................... IX
List of Symbols .................................................................................................... X
Chapter I. Introduction.......................................................................................... 1
I-1. What is Super-Resolution? ..................................................................... 1
I-2. Objectives ............................................................................................... 4
I-3. Outline .................................................................................................... 4
Chapter II. Literature Study .................................................................................. 5
II-1. The Formulation of Super Resolution................................................... 5
II-2. Keren Sub-Pixel Registration Method.................................................. 7
II-3. Scale-Invariant Feature Transform ....................................................... 9
II-3.1. Scale-Space Extrema Detection................................................ 10
II-3.2. Keypoint Localization .............................................................. 11
II-3.2.1. Interpolation of Nearby Data for Accurate Position ........ 12
II-3.2.2. Discarding Low-Contrast Keypoints................................ 12
II-3.2.3. Eliminating Edge Responses ............................................ 13
II-3.3. Orientation Assignment ............................................................ 14
II-3.4. Keypoint Descriptor.................................................................. 15
II-4. Speeded Up Robust Features .............................................................. 16
II-4.1. Point Detection – Fast Hessian Detector .................................. 17
II-4.2. Point Description – Haar-Wavelet Descriptor .......................... 20
II-4.3. Descriptor Matching – Fast Indexing with Laplacian Sign...... 22
II-
5. Random Sample Consensus ................................................................ 22
II-6. Normalized Convolution..................................................................... 24
II-6.1. Basis Functions and Applicability ............................................ 24
II-6.2. Definition.................................................................................. 25
II-6.3. Normalized Convolution Using Polynomial Bases.................. 26
II-7. Robust and Adaptive Normalized Convolution.................................. 29
II-7.1. Robust Normalized Convolution.............................................. 29
II-7.2. Structure-Adaptive Normalized Convolution........................... 30
II-7.3. Super Resolution Using Robust and Adaptive Normalized
Convolution ............................................................................... 34
Chapter III. Design Methodology ....................................................................... 36
III-1. SURF ................................................................................................. 36
III-1.1 Creating Integral Image............................................................ 37
III-1.2. Extracting Keypoints............................................................... 37
III-1.3. Generating the SURF Descriptor ............................................ 38
III-1.4. Matching.................................................................................. 39
III-2. Image Transformation........................................................................ 39
III-3. Image Reconstruction ........................................................................ 40
III-3.1. Certainty Optimization for Noise Robustness......................... 41
III-3.2. HR Reconstruction using Normalized Convolution ............... 41
III-3.3. Structure-Adaptive Normalized Convolution ......................... 42
Chapter IV. Experimental Results and Discussions............................................ 43
IV-1. Experiment Setup............................................................................... 43
IV-1.1. Simulation Platform................................................................. 43
IV-1.2. SURF Parameters .................................................................... 43
IV-1.3. LR Images................................................................................ 44
IV-1.4. Evaluation................................................................................ 48
IV-2. Experiment Results ............................................................................ 49
IV-2.1. Subjective Results ................................................................... 49
IV-
2.2. Objective Results..................................................................... 53
IV-3. Discussion.......................................................................................... 54
Chapter V. Conclusions and Future Works ......................................................... 58
V-1. Conclusions ......................................................................................... 58
V-2. Future Works ....................................................................................... 58
References…. ...................................................................................................... 59
Appendix A - Software Implementations…........................................................ 62
A-1. Initialization ........................................................................................ 62
A-2. SURF................................................................................................... 62
A-3. RANSAC ............................................................................................ 63
A-4. Normalized Convolution..................................................................... 63



List of Figures
Fig. I-1. Basic premise for super resolution.................................................... 3
Fig. II-1. Observation model relating LR images to HR images ..................... 6
Fig. II-2. SIFT Algorithm................................................................................. 9
Fig. II-3. Difference of Gaussian (DoG) ........................................................ 11
Fig. II-4. Detecting scale-space extrema of the scale normalized Laplacian 11
Fig. II-5. Compute the gradient of the pixels using 2D Gaussian kernel ...... 14
Fig. II-6. Weighted orientations histogram. ................................................... 15
Fig. II-7. Keypoint Descriptor........................................................................ 16
Fig. II-8. Integral image representation.......................................................... 17
Fig. II-9. The Gaussian partial derivatives..................................................... 19
Fig. II-10. Orientation assignment ................................................................... 20
Fig. II-11. Haar-wavelet responses for three different patterns ....................... 21
Fig. II-12. Fast Indexing with Laplacian sign.................................................. 22
Fig. II-13. Fast estimation of local scale by a quadratic interpolation along the
scale axis of a Gaussian scale-space of the HR density image....... 33
Fig. II-14. Robust and adaptive normalized convolution super-resolution
process............................................................................................. 34
Fig. III-1. The flowchart of the proposed method........................................... 36
Fig. III-2. SURF flowchart .............................................................................. 37
Fig. III-3. Up-scaling of the filter at constant image size................................ 38
Fig. III-4. Robust and Adaptive Normalized Convolution.............................. 40
Fig. III-5. Certainty Optimization for Noise Robustness ................................ 41
Fig. III-6. HR Reconstruction using Normalized Convolution....................... 42
Fig. III-7. Structure-Adaptive Normalized Convolution................................. 42
Fig. IV-1. Ballet ............................................................................................... 44
Fig. IV-2. Graffiti............................................................................................. 45
Fig. IV-3. Valley .............................................................................................. 46

Fig. IV-4. LR images generated from multiple images - Chart....................... 47
Fig. IV-5. LR images generated from multiple images - Dino ....................... 47
Fig. IV-6. Reconstructed HR images generated from 4 LR Ballet images using
different methods ............................................................................ 49
Fig. IV-7. Reconstructed HR images generated from 4 LR Graffiti images
using different methods................................................................... 50
Fig. IV-8. Reconstructed HR images generated from 4 LR Valley images using
different methods ............................................................................ 50
Fig. IV-9. Original and the reconstructed HR images generated from 4 LR
Chart images using different methods ............................................ 51
Fig. IV-10. Original and the reconstructed HR images generated from 4 LR
Dino images using different methods ............................................. 52
Fig. A-1. Select LR images dialogue box ...................................................... 62



List of Tables
Table IV-1. SURF Parameters............................................................................ 44
Table IV-2. Image rotation, shifting, and viewing angle difference parameters.48
Table IV-3. PSNR values (dB) of reconstructed HR images with resizing factor
2....................................................................................................... 53
Table IV-4. PSNR values (dB) of reconstructed HR images with resizing factor
4....................................................................................................... 54
Table IV-5. Computational time (seconds) needed to do image registration of
each method for HR image with resizing factor 2.......................... 56
Table IV-6. Computational Time (seconds) needed to do image registration of
each method for HR Image with resizing factor 4.......................... 57
References
[1] A. Katsaggelos, R. Molina, and J. Mateos, “Super Resolution of Images
and Video,” Synthesis Lectures on Image, Video, and Multimedia
Processing No. 7, Morgan & Claypool, 2007. [Online]. Available:
http://decsai.ugr.es/vip/files/books/SRbook.html
[2] S. C. Park, M. K. Park, and M. G. Kang, “Super Resolution Image
Reconstruction: a Technical Overview,” IEEE Signal Processing
Magazine, Vol. 20, No. 3, pp. 21-36, May 2003.
[3] R. Y. Tsai and T. S. Huang, "Multiframe Image Registration and
Restoration," Advances of Computer Vision and Image Processing, Vol. I,
pp. 317-339, 1984.
[4] S. Borman and R. L. Stevenson, “Spatial Resolution Enhancement of
Low-Resolution Image Sequences: A Comprehensive Review with
Directions for Future Research,” Laboratory for Image and Signal
Analysis (LISA), University of Notre Dame, Tech. Rep., July 1998.
[5] M. A. Fischler and R. C. Bolles, "Random Sample Consensus: A
Paradigm for Model Fitting with Applications to Image Analysis and
Automated Cartography," Comm. of the ACM, Vol. 24, pp. 381-395, June
1981.
[6] D. Keren, S. Peleg, and R. Brada, “Image Sequence Enhancement Using
Sub-pixel Displacement,” In Proc. IEEE Conf. on Computer Vision and
Pattern Recog., pp. 742-746, June 1988.
[7] D. G. Lowe, "Object Recognition from Local Scale-Invariant Features," In
Proc. of the Int. Conf. on Computer Vision, Vol. 2, pp. 1150-1157,
Sept .1999.
[8] Z. Yuan, P. M. Yan, and S. Li, “Super Resolution Based on Scale Invariant
Feature Transform,” In Proc of Int. Conf. on Audio, Language, and Image
Processing, pp. 1550-1554, July 7-9, 2008.
[9] H. Bay, T. Tuytelaars, and L. V. Gool, "SURF: Speeded Up Robust
Features," In Proc. of the 9th European Conf. on Computer Vision,
Springer LNCS, Vol. 3951, Part 1, pp. 404-417, 2006.
[10] T. Lindeberg, "Feature Detection with Automatic Scale Selection," Int. J.
of Computer Vision, Vol. 30, No. 2, pp. 79-116, 1998.
[11] K. G. Derpanis, “Integral Image-Based Representations,” Dept. Comp.
Science and Eng., York University, Tech. Memo, July 2007.
[12] K. G. Derpanis, “Overview of the RANSAC Algorithm,” Dept. Comp. Sci.
and Eng., York University, Tech. Memo, June 2005.
[13] H. Knutsson and C. F. Westin, “Normalized and Differential
Convolution,” In Proc. of IEEE Computer Society Conf. on Comp. Vision
and Pattern Recog., pp. 515-523, 1993.
[14] G. Farnebäck, Polynomial Expansion for Orientation and Motion
Estimation, Ph.D. dissertation, Department Elect. Eng., Linköping Univ.,
Sweden, Nov. 2002.
[15] T. Q. Pham, L. J. van Vliet and K. Schutte, “Robust Fusion of Irregularly
Sampled Data Using Adaptive Normalized Convolution,” EURASIP J. on
Applied Signal Processing, Vol. 2006, pp. 1-12, March 2006.
[16] C. Tomasi and R. Manduchi, “Bilateral Filtering for Gray and Color
Images,” In Proc. of IEEE Int. Conf. on Comp. Vision, pp. 839-846, Jan.
1998.
[17] M. Nitzberg and T. Shiota, “Nonlinear Image Filtering with Edge and
Corner Enhancement,” IEEE Trans. on PAMI, Vol. 14, No. 8, pp. 826-833,
1992.
[18] J. van de Weijer, L. J. van Vliet, P. W. Verbeek, and M. van Ginkel,
“Curvature Estimation in Oriented Patterns Using Curvilinear Models
Applied to Gradient Vector Fields,” IEEE Trans. on PAMI, Vol. 23, No. 9,
pp. 1035-1042, 2001.
[19] I. T. Young, L. J. van Vliet, and M. van Ginkel, “Recursive Gabor Filtering,” IEEE Trans. on Signal Processing, Vol. 50, No. 11, pp.
2798-2805, Nov. 2002.
[20] M. Irani and S. Peleg, “Improving Resolution by Image Registration,”
Graphical Models and Image Processing, Vol. 53, pp. 231-239, 1991.
[21] A. Neubeck and L. Van Gool, “Efficient Non-Maximum Suppression,” In
Proc. of Int. Conf. on Pattern Recog., Vol. 3, pp. 850 - 855, 2006.
[22] M. Brown and D. G. Lowe, “Invariant Features from Interest Point
Groups,” In Proc. of the 13th British Machine Vision Conf., pp. 253 - 262,
2002.
[23] D. Lingrand, “Particular Forms of Homography Matrices,” In Proc. of the
11th British Machine Vision Conf., Vol. 2, pp 596-605, Bristol (U. K.),
September 2000.
[24] A. Agarwal, C. V. Jawahar, and P. J. Narayanan, “A Survey of Planar
Homography Estimation Techniques,” Tech. Rep., Indian Institute of
Information Technology, June 2005.
[25] LCAV-Audiovisual Communications Laboratory, http://lcavwww.epfl.ch/
[26] Centre for Mathematical Sciences, http://www.maths.lth.se/
[27] SIFT for MATLAB, http://www.vlfeat.org/~vedaldi/code/sift.html
[28] Basic Image Processing Tutorial,
http://biocomp.cnb.csic.es/~coss/Docencia/ImageProcessing/Tutorial/inde
x.html
[29] P. Vandewalle, S. Süstrunk, and M. Vetterli, “A Frequency Domain
Approach to Registration of Aliased Images with Application to
Super-Resolution,” EURASIP J. on Applied Signal Processing, Vol. 2006,
pp. 1-14, March 2006.
QRCODE
 
 
 
 
 
                                                                                                                                                                                                                                                                                                                                                                                                               
第一頁 上一頁 下一頁 最後一頁 top
系統版面圖檔 系統版面圖檔