盲影像去模糊 (Blind Image Deblurring) 這個問題，在影像處理及信號處理的領域上已是存在相當長久的挑戰，此問題主要是希望能將因為拍照時因相機晃動而導致的模糊影像恢復為原有的清晰影像，但在影像去模糊的問題中只有模糊影像為已知，需要重建出的清晰影像以及造成影像模糊的模糊核皆為未知，因此在求解時需透過設定其他條件才能順利求解。
在本篇論文中我們提出了局部最大值先驗 (Patch-Wise Maximal Pixels Prior) 的先驗方法，我們發現在清晰的自然影像中，每個區塊內的最大值會相當大且趨近於1，但在模糊影像中則不會有此特性，因此可透過此條件作為先驗條件來處理影像去模糊的問題。
將我們所提出的局部最大值先驗與局部最小值先驗 (Patch-Wise Minimal Pixels Prior) 結合，成為局部極值先驗 (Local Extremums Prior) 的先驗方法，改善只使用局部最小值先驗時在亮度偏亮的影像會處理不佳的問題。
實驗結果顯示我們所提出的方法與目前其他現有的影像去模糊的方法在客觀的影像評估標準上皆可獲得差不多甚至更好的成績，並與過去相似的方法如暗通道先驗 (Dark Channel Prior) 及極通道先驗 (Extreme Channel Prior) 相比所需的計算時間大幅降低。

Blind image deblurring is a classical challenging problem in image processing and signal processing. This problem is trying to recover the blurred image which caused by the camera motion during taking picture. However, when trying to solve this problem, the information we only known is the blurred image, but we need to reconstruct two unknown information including clear image and blur kernel which causing image blurred. Therefore, when doing blind image deblurring problem, we need to have other additional information or constraints.
In this thesis, we present a new prior which called patch-wise maximal pixels prior. We found that in natural clear image, the maximal value in image patches would be very big and close to 1, but this characteristic would not show in blurred image. Therefore, we can use this observation as a prior to solve the blinding image deblurring problem.
To combine our patch-wise maximal pixels prior and patch-wise minimal pixels prior, it become the local extremums prior. Using local extremums prior can improve the problem that it has limitations when only use patch-wise minimal pixels prior while the input image is dominated by bright pixels.
Experimental results show that our method can have similar or even better results than state-of-the-art method. Also the runtime of our proposed method is shorter than the similar method like dark channel prior or extreme channel prior.

[1] R. Fergus, B. Singh, A. Hertzmann, S. T. Roweis, and W. T. Freeman, “Removing Camera Shake from a Single Photograph.” ACM Transactions on Graphics, vol. 25, no.3, pp. 787–794, 2006.
[2] A. Levin, Y. Weiss, F. Durand and W. T. Freeman, "Understanding and Evaluating Blind Deconvolution Algorithms," IEEE Conference on Computer Vision and Pattern Recognition, pp. 1964-1971, 2009.
[3] L. Xu and J. Jia, “Two-Phase Kernel Estimation for Robust Motion Deblurring,” European conference on computer vision, pp. 157–170, 2010.
[4] A. Levin, Y. Weiss, F. Durand and W. T. Freeman, "Efficient marginal likelihood optimization in blind deconvolution," IEEE Conference on Computer Vision and Pattern Recognition (CVPR), pp. 2657-2664, 2011.
[5] L. Xu, S. Zheng and J. Jia, "Unnatural L0 Sparse Representation for Natural Image Deblurring," IEEE Conference on Computer Vision and Pattern Recognition (CVPR), pp. 1107-1114, 2013.
[6] S. Cho and S. Lee, “Fast Motion Deblurring,” ACM Transactions on Graphics, vol. 28, no. 5, 2009, Art. no. 145.
[7] Q. Shan, J. Jia, and A. Agarwala, “High-quality Motion Deblurring from a Single Image,” ACM Transactions on Graphics, vol. 27, no. 3, 2008, Art. no. 73.
[8] J. Pan, Z. Hu, Z. Su and M. Yang, "Deblurring Text Images via L0-Regularized Intensity and Gradient Prior," IEEE Conference on Computer Vision and Pattern Recognition, pp. 2901-2908, 2014.
[9] J. Pan, D. Sun, H. Pfister and M. Yang, "Blind Image Deblurring Using Dark Channel Prior," IEEE Conference on Computer Vision and Pattern Recognition (CVPR), pp. 1628-1636, 2016.
[10] Y. Yan, W. Ren, Y. Guo, R. Wang and X. Cao, "Image Deblurring via Extreme Channels Prior," IEEE Conference on Computer Vision and Pattern Recognition (CVPR), pp. 6978-6986, 2017.
[11] L. Chen, F. Fang, T. Wang and G. Zhang, "Blind Image Deblurring With Local Maximum Gradient Prior," IEEE/CVF Conference on Computer Vision and Pattern Recognition (CVPR), pp. 1742-1750, 2019.
[12] F. Wen, R. Ying, Y. Liu, P. Liu and T. -K. Truong, “A Simple Local Minimal Intensity Prior and An Improved Algorithm for Blind Image Deblurring,” IEEE Transactions on Circuits and Systems for Video Technology, vol. 31, no. 8, pp. 2923-2937, 2021.
[13] Everingham, M., Van Gool, L., Williams, C. K. I., Winn, J. and Zisserman, A. “The PASCAL Visual Object Classes Challenge 2012 (VOC2012) Results,” 2012. http://www.pascal-network.org/challenges/VOC/voc2012/workshop/index.html.
[14] L. Xu, C. Lu, Y. Xu, and J. Jia, “Image Smoothing via L0 Gradient Minimization,” ACM Transactions on Graphics, vol. 30, no. 6, pp. 1-12, 2011.
[15] Zhou Wang, A. C. Bovik, H. R. Sheikh and E. P. Simoncelli, “Image Quality Assessment: From Error Visibility to Structural Similarity,” IEEE Transactions on Image Processing, vol. 13, no. 4, pp. 600-612, April 2004.
[16] R. Kohler, M. Hirsch, B. J. Mohler, B. Scholkopf, and S. Harmeling, “Recording and Playback of Camera Shake: Benchmarking Blind Deconvolution with a Real-World Database,” European conference on computer vision, pp. 27–40, 2012.