臺灣博碩士論文加值系統

在影像影像超解析度放大技術當中，希望將低解析度的影像或影片轉換成高解析度的模式。最主要會出現的問題就是放大的過程中出現模糊現象，影片動態模糊，取樣的錯誤，雜訊干擾等等。這些問題都常常被用來作為研究對象，在現階段有許多不錯的方法解決了以上的問題，試圖讓影像品質更好。其中一個較新的技術是稀疏編碼，此技術是一種壓縮感測方法並且用在影像重建中去除雜訊的影響[2]。在先前的技術當中，因為稀疏編碼的還原方式對於影像重建有一定品質的幫助，近幾年試圖解決影像公式處理中的錯誤訊號並且應用到影像超解析度放大技術。然而傳統的基於學習方式的影像超解析度放大技術需要大量的資料來輔助學習，所以應用上會需要尺寸大的字典來存放資料。雖然稀疏編碼已經將所有可能的資訊利用相較於傳統方法還少的基底組合創見出小型字典，但我們希望在重建影像的過程中還可以加速並且減少字典的大小。我們提出了顯著的方法去調整字典的大小並且同樣適合在傳統K-SVD演算法上作訓練產生出各式各樣不同大小的字典，並且加上K平均值分群法去解決計算複雜度的問題。將相同的影像區塊放在同一個類別當中，這樣的結果顯示讓觀看者可以有好的品質感受。另外我們也運用了結合梯度量和顯著圖區的新圖區透過每個影像區塊去選擇使用一般內插或稀疏編碼的方式讓影像還原過程可以加速。

Super-resolution reconstruction is the problem that we want to change the scale low-resolution images and videos to high-resolution. Frankly speaking, the main problem of super-resolution is eliminating the blurring effect like the motion blur, sampling errors, and noisy signal, etc. There are many ways to enhance the image resolution. One of the new algorithms is sparse coding algorithm which is the compressed sensing method and denoised the image reconstruction at first [2]. Since the sparse representation can represent the image patches well from the previous approaches, for recent years, the sparse coding solved the error signal based on the image formula process model and applied to single image super-resolution. But the conventional learning based super-resolution necessaries sufficient training data number of dictionary which means huge codebook size. Even the sparse coding only requires the dictionary with few atoms (basis) than the conventional approach but the testing time is long as the conventional approach. Therefore, we proposed a novel method to solve the single image super-resolution using the K-SVD method to generate the dictionary of the sparse coding with K-means clustering to overcome the problem of computational complexity. At learning phase, we gathered similar patches for proper class, the results shows the proposed method provides the fine quality for viewers. In addition, we apply the combination region of the gradient magnitude and saliency map to make the patch-based reconstruction step reduce the computation time.

Abstract i
摘 要 ii
Table of Contents iii
Chapter 1. Introduction 1
Chapter 2. Related Work 4
2.1. Interpolation Approach 4
2.2. Compressed Sensing 6
2.3. Saliency Detection 8
Chapter 3. Proposed Method 10
3.1. Feature Extraction 11
3.2. Saliency Detection Combination Approach 12
3.3. Sparse Coding Super-Resolution 14
3.3.1. Generic Image Reconstruction 14
3.3.2. Dictionary Size Reduction 16
3.4. Image Reconstruction Detail 19
Chapter 4. Experimental Results 20
Chapter 5. Conclusion and Future Work 38
References 39

[1] S. Baker and T. Kanade, “Limits on super-resolution and how to break them,” IEEE Trans. Pattern Analysis and Machine Intelligence, vol. 24, no. 9, pp. 1167-1183, 2002.
[2] M. Elad and M. Aharon, “Image Denoising Via Sparse and Redundant Representations Over Learned Dictionaries,” IEEE Trans. Image Processing, vol. 15, Issue. 12, pp. 3736-3745, 2006.
[3] M. Irani and S. Peleg, “Motion analysis for image enhancement: resolution, occlusion and transparency,” Journal Visual Communication and Image Representation, vol. 4, no. 4, pp. 324-335, 1993.
[4] J. D. Van Ouwerkerk, “Image super-resolution survey, ” in Proc. Int. Journal Image and Vision Computing, vol. 24, no. 10, pp. 1039-1052, 2006.
[5] R. Keys, “Cubic convolution interpolation for digital image processing,” IEEE Trans. Acoustics, Speech and Signal Processing, vol. 29, no. 6, pp. 1153-1160, 1981.
[6] J. Sun, Z. Xu, and H. Y. Shum, “Gradient profile prior and its applications in image super-resolution and enhancement,” IEEE Trans. Image Processing, vol. 20, Issue. 6, pp. 1529-1542, 2011.
[7] W. T. Freeman, T. R. Jones, and E.C. Pasztor, “Example-based super-resolution,” IEEE Trans. Computer Graphics and Applications, vol. 22, no. 2, pp. 55-65, 2002.
[8] W.T. Freeman, E. Pasztor, and O. Carmichael, “Learning low-level vision,” in Proc. Int. Conf. Computer Vision, vol. 2, 1999.
[9] Q. Shan, Z. Li, J. Jia, and C.-K. Tang, “Fast image/video upsampling,” ACM Trans. Graphics, vol. 27, no. 5, 2008.
[10] J. Yang, J. Wright, T. Huang, and Y. Ma, “Image super-resolution via sparse representation,” IEEE Trans. Image Processing, vol. 19, Issue. 11, pp. 2861-2873, 2010.
[11] M. Aharon, M. Elad, and A. Bruckstein, “K-SVD: An algorithm for designing overcomplete dictionaries for sparse representation,” IEEE Trans. Signal Processing, vol. 54, no. 11, pp. 4311–4322, Nov. 2006.
[12] X. D. Hou, and L. Q. Zhang, “Saliency detection: a spectral residual approach,” in Proc. IEEE Int. Conf. Computer Vision and Pattern Recognition, pp. 1-8, 2006.
[13] C. L. Guo, Q. Ma, and L. M. Zhang, “Spatio-temporal saliency detection using phase spectrum of quaternion fourier transform,” in Proc. IEEE Int. Conf. Computer Vision and Pattern Recognition, pp. 1-8, 2008
[14] J. Sun, J. Sun, Z. B. Xu, and H. Y. Shum, “Image super-resolution using gradient profile prior,” in IEEE Conf. Computer Vision and Pattern Recognition, pp. 1-8, 2008.
[15] R. Tibshirani, “Regression shrinkage and selection via the lasso,” in J. Roy. Statist. Soc., Ser. B, vol. 58, no. 1, pp. 267–288, 1994.
[16] Z. Wang, A. C. Bovik, H. R. Sheikh and E. P. Simoncelli, “Image quality assessment: From error visibility to structural similarity,” IEEE Trans. Image Processing, vol. 13, no. 4, pp. 600-612, pp. 600–612, 2004.