Weisheng Dong, Lei Zhang, Guangming
Shi, and Xin Li
¡°Nonlocally Centralized Sparse Representation for Image
Restoration,¡±
IEEE Trans. on Image Processing, vol. 22, no. 4,
pp. 1620-1630, Apr. 2013.
Paper: download
here
Code: download here
Part A: Results on Image Denoising
Notes:
(1)
The
denoising method in [1] is labeled as ¡°SAPCA-BM3D¡±;
(2)
The
denoising method in [2] is labeled as ¡°LSSC¡±;
(3)
The
denoising method in [3] is labeled as ¡°EPLL¡±;
(4)
The
proposed denoising method is labeled as ¡°NCSR¡±.
For example, the denoised image by the method NCSR on image Girl is labeled as ¡°NCSR_Girl¡±. Other denoised images are labeled similarly.
The
denoising results on Lena |
|
The
denoising results on Monarch |
|
The denoising results on Barbara |
|
The denoising results on Boat
|
|
The denoising results on C.
Man
|
|
The denoising results on Couple
|
|
The denoising results on F.
Print
|
|
The denoising results on Hill
|
|
The denoising results on House
|
|
The denoising results on Man
|
|
The denoising results on Peppers |
|
The
denoising results on Straw
|
Part B: Results on Image Deblurring
Notes:
(1)
The
deblurring method in [4] is labeled as ¡°FISTA¡±;
(2)
The
deblurring method in [5] is labeled as ¡°l0-sparse¡±;
(3)
The
deblurring method in [6] is labeled as ¡°IDD-BM3D¡±;
(4)
The
deblurring method in [7] is labeled as ¡°ASDS-Reg¡±;
(5)
The
deblurring method in [8] is labeled as ¡°Fergus¡±;
(6)
The
deblurring method in [11] is labeled as ¡°TVMM¡±;
(7)
The
proposed denoising method is labeled as ¡°NCSR¡±.
For example, the deblurred image by the method NCSR on image
Butterfly is labeled as ¡°NCSR_Butterfly¡±. Other deblurred images are labeled
similarly.
Experiment 1: 9¡Á9 uniform blur kernel, noise level
Some additional deblurring results by the proposed NCSR method with different parameters: Download
images
The
deblurring results on Butterfly
|
|
The
deblurring results on Boats
|
|
The deblurring results on Cameraman
|
|
The deblurring results on House
|
|
The deblurring results on Parrot
|
|
The deblurring results on Lena
|
|
The deblurring results on Barbara
|
|
The deblurring results on Starfish
|
|
The deblurring results on Peppers
|
|
The
deblurring results on Leaves
|
Experiment 2: Gaussian blur kernel (standard deviation 1.6),
noise level
Some additional deblurring results by the proposed NCSR method with different parameters: Download
images
The
deblurring results on Butterfly
|
|
The
deblurring results on Boats
|
|
The deblurring results on Cameraman
|
|
The deblurring results on House
|
|
The deblurring results on Parrot
|
|
The deblurring results on Lena
|
|
The deblurring results on Barbara
|
|
The deblurring results on Starfish
|
|
The deblurring results on Peppers
|
|
The
deblurring results on Leaves
|
Experiment 3: Motion deblurring
The
deblurring results on Oldman
|
|
The deblurring results on lyndsey2
|
|
The
deblurring results on Test7
|
Experiment
4:
6 typical deblurring experiments presented in [6]
The
deblurring results on Cameraman256
|
|
The deblurring results on House
|
|
The
deblurring results on Lena512
|
|
The deblurring results on Barbara |
Part C: Results on Image Super-resolution
Notes:
(1)
The
super-resolution method in [9] is labeled as ¡°TV¡±;
(2)
The
super-resolution method in [10] is labeled as ¡°Sparse¡±;
(3)
The
super-resolution method in [7] is labeled as ¡°ASDS-Reg¡±;
(4)
The
proposed super-resolution method is labeled as ¡°NCSR¡±.
For example, the high resolution (HR) image reconstructed by
the method NCSR on image Girl is
labeled as ¡°NCSR_Girl¡±. Other super-resolution
results labeled similarly.
Experiment 1: Super-resolution on noiseless images
The
HR images with scalar factor 3 on Butterfly
|
|
The
HR images with scalar factor 3 on Flower
|
|
The HR images with scalar
factor 3 on Girl
|
|
The HR images with scalar
factor 3 on Parthenon
|
|
The HR images with scalar
factor 3 on Parrot
|
|
The HR images with scalar
factor 3 on Raccoon
|
|
The HR images with scalar
factor 3 on Bike
|
|
The HR images with scalar
factor 3 on Hat
|
|
The
HR images with scalar factor 3 on Plants
|
Experiment 2: Super-resolution on noisy images
Gaussian white noise with
standard deviation 5 is added to the LR images to simulate the noisy low
resolution images.
The
HR images with scalar factor 3 on Butterfly
|
|
The
HR images with scalar factor 3 on Flower
|
|
The HR images with scalar
factor 3 on Girl
|
|
The HR images with scalar
factor 3 on Parthenon
|
|
The HR images with scalar
factor 3 on Parrot
|
|
The HR images with scalar
factor 3 on Raccoon
|
|
The HR images with scalar
factor 3 on Bike
|
|
The HR images with scalar
factor 3 on Hat
|
|
The
HR images with scalar factor 3 on Plants
|
References
[1]
K. Dabov, A. Foi, V. Katkovnik, and K. Egiazarian,
¡°BM3D image denoising with
shape-adaptive principal component analysis,¡± Proc. Workshop on Signal Processing with Adaptive Sparse Structured
Representation (SPARS¡¯09), Saint-Malo, Fance,
Apr. 2009.
[2]
J. Mairal, F. Bach, J. Ponce,
G. Sapiro and A. Zisserman,
¡°Non-Local Sparse Models for Image Restoration,¡± in Proc. IEEE International
Conference on Computer Vision, Tokyo, Japan, 2009.
[3]
D. Zoran and Y. Weiss,
¡°From learning models of natural image patches to whole image restoration,¡± in Proc. IEEE Int. Conf. on Computer Vision
(ICCV), 2011.
[4]
A.
Beck and M. Teboulle, ¡°Fast gradient-based algorithms
for constrained total variation image denoising and deblurring problems,¡± IEEE Trans. on
Image Process., vol. 18,
no. 11, pp. 2419-2434, Nov. 2009.
[5]
J. Portilla, ¡°Image
restoration through L
[6]
A. Danielyan, V. Katkovnik, and K. Egiazarian, ¡°BM3D frames and variational image deblurring,¡±
IEEE Trans. on Image Processing, vol. 21, no. 4, Apr. 2012.
[7]
W. Dong, L. Zhang, G. Shi, and X. Wu, ¡°Image
deblurring and super-resolution by adaptive sparse domain selection and
adaptive regularization,¡± IEEE Trans. on
Image Processing, vol. 20, no. 7, pp. 1838-1857, July 2011.
[8]
R. Fergus, B. Singh, A. Hertzmann,
S. T. Roweis, and W. T. Freeman, ¡°Removing camera
shake from a single image,¡± ACM Trans.
Graph. (SIGGRAPH), pages 787-794, 2006.
[9]
A. Marquina, and S. J. Osher, ¡°Image super-resolution by
TV-regularization and Bregman iteration,¡± J. Sci. Comput., vol.
37, pp. 367-382, 2008.
[10] J. Yang, J. Wright, Y. Ma, and T. Huang,
¡°Image super-resolution as sparse representation of raw image patches,¡± IEEE Computer Vision and Pattern
Recognition, vol. 1, pp. 1-8, Jun. 2008.
[11] J. Oliveira, J. M. Bioucas-Dias,
and M. Figueiredo, ¡°Adaptive
total variation image deblurring: a majorization-minimization
approach,¡± Signal Processing, vol.
89, no. 9, pp. 1683-1693, Sep. 2009.