臺灣博碩士論文加值系統

由於影像顯示裝置的進步，人們對影像的品質要求越來越高。但是高級的影像擷取設備卻依然較為昂貴，所以我們常轉而採用軟體方式來加強影像品質。超解析度演算法即為當前十分熱門的影像處理研究領域之一，應用的範圍很廣，包括軍事、監視系統等等。
近年來，線上數位學習的需求越來越高，但是大部分的教師並沒有較好的影像擷取設備，故無法錄製高品質的教學內容。本篇論文即是針對一般教學的影像加以強化，特別是以黑板、白板或投影幕為教具的教學方式。
超解析度演算法雖然可以增強影像解析度，但是大量的運算時間和未參考影像本身的內容是大部分超解析度演算法所擁有的問題。在本篇論文中，使用了邊緣偵測來估計小區塊之中影像邊緣的稠密程度，並使用平均移動法來進行色彩分割。綜合上述兩種資訊來判斷何處是教學影像中，人眼最關注、次關注及未關注的區域，並在不同的區域使用不同複雜度的演算法。
實驗結果顯示，我們僅需使用高複雜度演算法來處理全圖約20%的區塊，大幅降低運算時間，同時也可以得到品質良好的影像輸出。

Because of the development of image display devices, people require much better quality of images. The high-level image capture devices are still expensive, therefore we usually use software to enhance the quality of images. Super- resolution algorithm is a popular research domain in digital image processing, and the applications are widespread, including the military, surveillance, and so on.
Recently, the requirement of on-line digital learning is much higher, but many teachers do not have good image capture devices, they cannot record teaching contents with high quality. In this paper, we propose a method to enhance the normal teaching images, especially for the teaching mode using black board, white board, or projection screen.
Though super-resolution algorithm can enhance the image resolution, the large execution time and disregarding the image content are the problems in majority of super-resolution algorithm. In this paper, we use edge detection to estimate the image edge density of small blocks, and use mean shift to implement color segmentation. We can integrate the above information to determine where people pay attention to mostly, where secondly, and where we do not care, and use different complexity algorithms to process them.
By the experiment result, we only have to process 20% area of the whole image, and decrease execution time significantly. Simultaneously, we can only get an image output with good quality.

誌 謝 i
圖 錄 v
表 錄 vi
Chapter 1 Introduction 1
1.1 Motivation 1
1.2 Introduction to Super Resolution 2
Chapter 2 Related Works 4
2.1 Interpolation Method 4
2.1.1 Nearest Neighbor Method 4
2.1.2 Bilinear Interpolation 5
2.1.3 Bi-Cubic Interpolation 6
2.1.4 Edge-Directed Interpolation 7
2.2 Projection onto Convex Sets 8
2.3 Iterative Back-Projection (IBP) 9
2.4 Maximum a Posterior (MAP) 11
2.5 Image Segmentation by Mean Shift 13
2.5.1 Mean Shift 13
2.5.2 Image Segmentation 15
2.6 Edge Detection 17
2.6.1 Sobel Operator 18
2.6.2 Laplacian of Gaussian Filter 19
Chapter 3 Proposed Method 21
3.1 Flow Chart 21
3.2 Major Color Estimation 23
3.3 Edge Density Estimation 25
3.4 Combination 27
3.4.1 Unsharp Mask 27
3.4.2 Combination 28
3.4.3 Deblock 30
Chapter 4 Experiments and Results 32
4.1 Experimental Environment 32
4.2 Experiment Result 33
Chapter 5 Conclusion and Future Work 49
5.1 Conclusion 49
5.2 Future Work 49
Reference 50
圖 錄
Figure 1.1 Basic premise for super resolution [9]. 2
Figure 2.1 Nearest neighbor method. 4
Figure 2.2 Bilinear interpolation. 5
Figure 2.3 Bi-cubic convolution. 6
Figure 2.4 Edge directed interpolation. 7
Figure 2.6 Observation model relating LR images to HR images [9]. 9
Figure 2.7 Illustration of IBP method [9]. 10
Figure 2.8 (a) Original digitized airport image. (b) Image expanded by the zero-order hold method. (c) By linear interpolation. (d) By MAP estimation technique [10]. 12
Figure 2.9 (a) A 400*276 color image. (b) Corresponding L*u*v color space with 110,400 data points. (c) 2-dimensional data set of the first two component of (b). (d) Decomposition obtained by the mean shift procedure. (e) The topography corresponding to the density estimation [2]. 14
Figure2.10 (a) Input. (b) Mean shift paths for the pixels on the plateau and on the line. (c) Filtering result (hs, hr) = (8, 4). (d) Segmentation result [2]. 15
Figure 2.11 Original Baboon image and segmentation with different hs and hr [2] 16
Figure 2.12 Finding the different derivatives. 17
Figure 2.13 (a) Original image. (b) Edge detection by Sobel operator. (c) Edge detection by Laplacian filter without Gaussian smoothing. (d) Edge detection by Laplacian filter with Gaussian smoothing. 20
Figure 3.1 The flow chart of our method. 21
Figure 3.1 Common teaching image examples. 22
Figure 3.2 Results of down-sampling and color segmentation. 23
Figure 3.3 Major color estimation results, the brighter regions are major color. 24
Figure 3.4 Results of Sobel operator. 25
Figure 3.5 Results of edge density estimation. 27
Figure 3.6 (a) The original image. (b) The blur image by Gaussian filter. (c) The difference between (a) and (b) by substraction. (d) The sharp image by adding (c) to (a). 28
Figure 3.7 Combination of edge density estimation result and major color estimation result. 29
Figure 3.8 Results of E∩C. 30
Figure 3.9 (a) The output image before deblock. (b) The result after deblock. 31
Figure 4.1 The broken line graph of PSNR. Ordinate axis represents dB. 47
Figure 4.2 The bar chart of execution time. Ordinate axis represents seconds. 48
表 錄
Table 4.1 Statistics of block classification. 33
Table 4.2 Statistics of PSNR in dB. 47
Table 4.3 Statistics of execution time in second. 48

[1]J. Allebach and P. W. Wong, “Edge-Directed Interpolation,” Proceedings of IEEE Int. Conf. on Image Processing, Lausanne, Switzerland, pp. 707-710, 1996.
[2]D. Comaniciu, P. Meer, “Mean shift: a robust approach toward feature space
analysis,” IEEE Trans. Pattern Anal. Mach. Intell., Vol. 24, No. 5, pp.603–619, 2002
[3]M. Elad and A. Feuer, “Restoration of a single super resolution image from several blurred, noisy, and under sampled measured images,” IEEE Trans. Image Processing, vol. 6, no. 12, pp. 1646-1658, 1997.
[4]S. Farsiu, M. D. Robinson, M. Elad, and P. Milanfar, “Fast and Robust Multi-Frame Super Resolution,” IEEE Transactions on Image Processing, Vol. 13, No. 10, pp. 1327–1344, 2004.
[5]R.C. Gonzalez, and R.E. Woods, Digital Image Processing. Addison-Wesley
Pub Co, 2nd Edition, 2002.
[6]T. S. Huang and R. Y. Tsai, “Multi-Frame Image Restoration and Registration,” Adv. Comput. Vis. Image Process., Vol. 1, pp. 317–339, 1984.
[7]M. Irani and S. Peleg, “Improving Resolution by Image Registration,” CVGIP: Graphical Models and Image Proc., Vol. 53, pp. 231-239, 1991.
[8] R. G. Keys, “Cubic Convolution Interpolation for Digital Image Processing,” IEEE Trans. Acoustics, Speech, and Signal Processing, Vol. 29, No. 6, pp. 1153-1160, 1981.
[9]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, 2003.
[10]R. R. Schultz and R. L. Stevenson, “A Bayesian Approach to Image Expansion for Improved Definition,” IEEE Transactions on Image Processing, Vol. 3, No. 3, pp. 233-242, 1994.
[11] H. Stark and P. Oskoui, “High Resolution Image Recovery from Image-Plane Arrays, Using Convex Projections,” J. Opt. Soc. Am. A, Vol. 6, pp. 1715-1726, 1989.
[12]Wikipedia, “Interpolation,” http://en.wikipedia.org/wiki/Interpolation, 2011.
[13] Wikipedia, “Gaussian blur,” http://en.wikipedia.org/wiki/Gaussian_blur, 2011.