Local Approximations in Signal and Image Processing

EXPERIMENTAL RESULTS

In this page are collected some of the results (images and tables) which can be obtained using the LASIP routines for anisotropic nonparametric image restoration.

Further experimental results are reported in the publications, which can be downloaded from the main page.

Denoising

Denoising (from additive Gaussian white noise with variance σ²) results (PSNR, dB) for the Anisotropic LPA-ICI Recursive denoising technique:

σ	Lena 512x512	Boats 512x512	Cameraman 256x256	House 256x256	Peppers 512x512	Peppers 256x256	Barbara 512x512
5	37.93	36.17	37.74	38.09	37.07	37.60	36.50
10	34.56	32.74	33.35	34.93	34.38	33.87	31.97
15	32.72	30.87	31.09	33.18	32.87	31.78	29.44
20	31.44	29.58	29.74	31.82	31.77	30.30	27.75
25	30.43	28.58	28.68	30.73	30.84	29.16	26.50
30	29.60	27.77	27.76	29.76	30.04	28.18	25.58
35	28.90	27.08	26.94	28.90	29.31	27.36	24.89
50	27.23	25.52	24.92	26.86	27.48	25.41	23.60

These results can be reproduced using the routine demo_RecursiveDenoisingGaussian.m from the LASIP software package, which implements the recursive anisotropic denoising method presented in:
PDF

Foi, A., V. Katkovnik, K. Egiazarian, and J. Astola, “A novel anisotropic local polynomial estimator based on directional multiscale optimizations”, Proc. of the 6th IMA Int. Conf. Math. in Signal Processing, Cirencester (UK), pp. 79-82, 2004.

Deblurring

Deblurring results (ISNR, PSNR, dB) using the Anisotropic LPA-ICI Regularized-Wiener Inverse deconvolution:

Experiment #	1	2	3	4
ISNR	8.29	7.86	6.00	3.95
PSNR	29.06	30.09	28.17	32.44

Experiment 1 - Cameraman 256x256, blur: 9x9 uniform "box-car" blur, Gaussian white noise: BSNR 40dB
Experiment 2 - Cameraman 256x256, blur: 15x15 1/(x²+y²), x,y=-7,...,7, kernel, Gaussian white noise: σ²=2
Experiment 3 - Cameraman 256x256, blur: 15x15 1/(x²+y²), x,y=-7,...,7, kernel, Gaussian white noise: σ²=8
Experiment 4 - Lena 512x512, blur: 5x5 separable [1, 4, 6, 4, 1]/16 filter, Gaussian white noise: BSNR 15.93dB

Images corresponding to the above table are also available.

These results can be reproduced using the routine demo_DeblurringGaussian.m from the LASIP software package, which implements the anisotropic deconvolution method presented in:
PDF

Katkovnik, V., A. Foi, K. Egiazarian, and J. Astola, “Directional varying scale approximations for anisotropic signal processing”, Proc. of XII European Signal Process. Conf., EUSIPCO 2004, pp. 101-104, 2004.

Inverse Halftoning

Inverse Halftoning results (PSNR, dB) for the Anisotropic LPA-ICI Inverse Halftoning (based on the Regularized-Wiener Inverse deconvolution):

error-diffusion	Lena 512x512	Boats 512x512	Peppers 512x512
Floyd et al.	32.54	29.63	31.83
Jarvis et al.	33.10	30.13	31.70

Images corresponding to the above table are also available.

These results can be reproduced using the routine demo_InverseHalftoning.m from the LASIP software package, which implements the anistropic inverse halftoning method presented in:
PDF

Foi, A., V. Katkovnik, K. Egiazarian, and J. Astola, “Inverse halftoning based on the anisotropic LPA-ICI deconvolution”, Proc. Int. TICSP Workshop Spectral Meth. Multirate Signal Process., SMMSP 2004, Vienna, pp. 49-56, 2004.