文件列表:

l0tv (0, 2014-11-21)
l0tv\cimg (0, 2014-11-21)
l0tv\cimg\barbara.png (181026, 2014-06-07)
l0tv\cimg\blonde.png (164651, 2014-06-08)
l0tv\cimg\boat.png (155189, 2014-06-07)
l0tv\cimg\cameraman.png (116068, 2014-06-07)
l0tv\cimg\house.png (83935, 2014-06-08)
l0tv\cimg\jetplane.png (138669, 2014-06-08)
l0tv\cimg\jetplane_color.png (425237, 2014-11-10)
l0tv\cimg\lake.png (171540, 2014-06-08)
l0tv\cimg\lenna.png (154036, 2013-09-05)
l0tv\cimg\lenna_color.png (474350, 2014-06-06)
l0tv\cimg\livingroom.png (160760, 2014-06-08)
l0tv\cimg\mandrill.png (207015, 2014-06-07)
l0tv\cimg\pepper.png (55856, 2014-06-07)
l0tv\cimg\pepper_color.png (507582, 2014-11-10)
l0tv\cimg\pirate.png (176400, 2014-06-07)
l0tv\cimg\walkbridge.png (186501, 2014-06-08)
l0tv\demo.m (1494, 2014-11-21)
l0tv\demo_generate_corrupted_images.m (2762, 2014-11-21)
l0tv\dimg (0, 2014-11-21)
l0tv\result (0, 2014-11-21)
l0tv\result\demo_write_latex_deblurring.m (3054, 2014-11-20)
l0tv\result\demo_write_latex_denoising.m (2820, 2014-11-20)
l0tv\result\test_admm_l1tv.mat (206322, 2014-06-29)
l0tv\result\test_admm_l1tv.txt (177455, 2014-06-29)
l0tv\result\test_amf.mat (8589, 2014-11-08)
l0tv\result\test_amf.txt (15503, 2014-11-08)
l0tv\result\test_aop_l02tv.mat (63239, 2014-06-28)
l0tv\result\test_aop_l02tv.txt (63283, 2014-06-28)
l0tv\result\test_direct_admm_l0tv.mat (200680, 2014-07-04)
l0tv\result\test_direct_admm_l0tv.txt (173810, 2014-07-04)
l0tv\result\test_l1tv.mat (250627, 2014-11-08)
l0tv\result\test_l1tv.txt (391416, 2014-11-08)
l0tv\result\test_mpec_padmm_l0tv.mat (201242, 2014-07-03)
l0tv\result\test_mpec_padmm_l0tv.txt (173646, 2014-07-03)
l0tv\result\test_pd_l0tv.mat (200495, 2014-06-23)
l0tv\result\test_pd_l0tv.txt (173735, 2014-06-23)
l0tv\scratched (0, 2014-11-21)
... ...

l0TV README File ================================================= 1. Quick-start To test the code, we have provided a demo Matlab script "demo.m", which runs l0TV_PADMM over the "barbara.png" image. Simply type "demo" in the Matlab command prompt. If everything is working properly, you will get the following results: (1) SNR of corrupted image. SNR0:0.07, SNR1:0.19, SNR2:-0.55. (2) SNR of recovered image. SNR0:0.09, SNR1:1.58, SNR2:4.30. 2. Using the code To reproduce the experiments which described in [1], you need to perform the following two steps: (1) run "demo_generate_corrupted_images.m" to generate the corrupted images. (2) run Matlab script "test0_*.m / test1_*.m / test2_*.m / test3_*.m / test4_*.m" to obtain the experiment results. 3. Directory tree We demonstrate the directory tree of our code, including the names of some main functions and their descriptions. ------------------- solvers | | | |-------- l0tv_padmm_color.m : Implementation of PADMM algorithm for solving Colored l0TV problem | |-------- l0tv_proj_reg.m : Implementation of the penalty decomposition method and Augmented Lagrangian method | |-------- tvinpaint.m : Implementation of the Split Bregman Method for l1TV and l2TV inpaint problems | |-------- l1tv_admm.m : Implementation of the proximal ADMM for solving l1TV | |-------- amf.m : Implementation of the adaptive median filter to salt-and-pepper inpulse denoising | |-------- acwmf2.m : Implementation of the adaptive center-weighted median filter to random-valued inpulse denoising | | ------------------- util | | | |-------- snr_l0.m : compute the SNR0 value for images | |-------- snr_l1.m : compute the SNR1 value for images | |-------- snr_l2.m : compute the SNR2 value for images | |-------- GenBlurOper.m : Generate the blurring kernel | |-------- threadholding_l0.m : proximal operator for the l0 norm | |-------- functionAX.m : specify the matrix-vector for the denoising/deblurring problem | |-------- impulsenoise.m : inject an image with (Salt-and-pepper or Random-valued) impulse noise | |-------- difX.m/difY.m : Given u, these functions compute \nabla_x u and \nabla_y u, respectively. | |-------- divX.m/divY.m : Given z, these functions compute - \nabla_x^T z and - \nabla_y^T z, respectively. | | ------------------- cimg | | | |-------- Clean images | | ------------------- dimg | | | |-------- Dirty images | | ------------------- scratched | |-------- Scratched images 4. REFERENCES: [1] Ganzhao Yuan, Bernard Ghanem. $\ell_0$TV: A New Method for Image Restoration in the Presence of Impulse Noise. Oral Presentation, CVPR 2015.

近期下载者：

相关文件：

评论：[我要评论] [举报此文件]

收藏者：