Image Deblurring Using Dark Channel Prior Liang Zhang (lzhang432)
Motivation Solutions Dark Channel Model Optimization Application Future Work Reference Outline
Motivation Recover Blur Image Photos are taken everyday (mobile phone, digital camera, GoPros) Blur Images are undesirable Hard to reproduce the capture moment How to get deblur image without have to retake picture? Example of Blur Image
Motivation Blur Image Clear Image Blur Kernel Noise Blur Image = Sharp Image * Blur Kernel + Noise[8]
Motivation Blur Image Clear Image Blur Kernel Noise Blur Image = Sharp Image * Blur Kernel + Noise[8]
Motivation Solutions Dark Channel Model Optimization Application Future Work Reference Outline
Convolution Convolution: Weighted average of a patch 50 10 0 120 30 0 180 25 90 60 90 2 20 60 60 75 60 3 80 20 0 15 5 0 20 15 0 20 25 0 12 300 200 0 0 150 1/9 1/9 1/9 1/9 1/9 1/9 1/9 1/9 1/9 Blur Kernel Clear Image
Convolution Convolution: Weighted average of a patch 50 10 0 120 30 0 180 25 90 60 90 2 20 60 60 75 60 3 80 20 0 15 5 0 20 15 0 20 25 0 12 300 200 0 0 150 1/9 1/9 1/9 1/9 1/9 1/9 1/9 1/9 1/9 Blur Kernel 45 Clear Image
Convolution Convolution: Weighted average of a patch Dark Channel: The lowest value among a patch 50 10 0 120 30 0 180 25 90 60 90 2 20 60 60 75 60 3 80 20 0 15 5 0 20 15 0 20 25 0 12 300 200 0 0 150 1/9 1/9 1/9 1/9 1/9 1/9 1/9 1/9 1/9 Blur Kernel 45 >= 0 Clear Image
Convolution and Dark Channel Dark Channel Image Dark channel of blurred image are less sparse than the dark channel of sharp image Clear Image Blurred Image
Deblur Approach Dark Channel Image Clear Image Blurred Image
Motivation Solutions Dark Channel Model Optimization Application Future Work Reference Outline
Model Based of fact that the dark channel of sharp image have more number of zerointensity pixels [8]
Model Based of fact that the dark channel of sharp image have more number of zerointensity pixels [8] Data Fitting
Model Based of fact that the dark channel of sharp image have more number of zerointensity pixels Blur Kernel Regulari zation [8]
Model Based of fact that the dark channel of sharp image have more number of zerointensity pixels Gradients of Image Sparsity [8]
Model Based of fact that the dark channel of sharp image have more number of zerointensity pixels [8] L0 norm Nonlinear min operator This term is used to measure the sparsity of dark channel
Non-linear Operation D(I) = MI [8] D(I): vectorized of D(I) M: indicator matrix of dark channel I: vectorized latent image M Latent Image Dark Channel
Non-linear Operation D(I) = MI [8] D(I): vectorized of D(I) M: indicator matrix of dark channel I: vectorized latent image M Latent Image Dark Channel
Non-linear Operation D(I) = MI [8] D(I): vectorized of D(I) M: indicator matrix of dark channel I: vectorized latent image M Latent Image Dark Channel
Blur Kernel and Clear Image Dark Channel of Latent Image Blur Kernel Clear Image
Motivation Solutions Dark Channel Model Optimization Application Future Work Reference Outline
Application Run with paper s data set
Application Run with paper s data set
Application Run with our own data set (blur images are download from google)
Motivation Solutions Dark Channel Model Optimization Application Future Work Reference Outline
Future Work Parallel computing implementation to accelerate deblur computing Implement the dark channel on mobile phone Improve the dark channel methods
Motivation Solutions Dark Channel Model Optimization Application Future Work Reference Outline
Reference [1] L.B.Lucy. An iterative technique for the rectification of observed distributions. Astronomy Journal, 79(6):745-754, 1974. [2] T, Chan anc C.Wong. Total variation blind deconvolution.. IEEE TIP, 7(3):370-375, 1998. [3] R.Fergus, B.Singh, A.Hertzmann, S.T.Roweis, and W.T.Freeman. Removing camera shake from a single photograph. ACM SIGGRAPH, 25(3): 787-794, 2006.
Reference [4] Y.Hacohen, E. Shechtman, and D.Lischinski. Deblurring by example using dense correspondence. In ICCV, pages2384-2391, 2013 [5] D.Krishnan, T.Tay, and R.Fergus. Blind deconvolution using a normalized sparsit measure. In CVPR, pages2657-2664, 2011 [6] T.Michaeli and M.Irani Blind deblurring using internal patch recurrence. In ECCV, pages783-798, 2014
Reference [7] K.He, J.Sun, and X.Tang Single image haze removal using dark channel prior. In CVPR, pages1956-1963, 2009 [8] Jinshan Pan, Deqing Sun, Hanspeter Pfister, and Ming-Hsuan Yang Blind Image Deblurring Using Dark Channel Prior. In CVPR, 2016 [9]https://sites.google.com/site/jspanhomepage/ [10]https://www.youtube.com/watch?v=dl1_592iDUY [11]https://www.google.com/search? q=blur+image&espv=2&source=lnms&tbm=isch&sa=x&ved=0ahukew jfntk7vrztahuo_imkhrnfam4q_auibigb&biw=1074&bih=709&dpr=1
Thanks
Q & A
Backup slides