Local Approximations in Signal and Image Processing

Anisotropic Nonparametric Image Restoration DemoBox

for MATLAB version 7.5 or later


The LASIP Anisotropic Nonparametric Image Restoration DemoBox is a set of MATLAB routines for image restoration (denoising, deblurring, inverse-halftoning, etc.).
They implement a recent new development in the area of statistical scale-adaptive local approximation techniques. LASIP provides flexible tools for the design of filters equipped with scale (window size) parameters. Directional filters can also be designed. The adaptivity of these filters is enabled by special statistical rules for a pointwise-adaptive selection of the scale values. The multidirectional versions these filters are especially efficient for anisotropic data.

The main algorithms are prepared as demos, so that they can be executed in a straightforward manner.
These demos reproduce figures and results from the publications by the authors of the LASIP project and their collaborators.

All the provided demos are open-source, and may be modified and tuned to be exploited with other data.

This DemoBox is available free-of-charge for educational and non-profit scientific research, enabling others researchers to understand and reproduce our work.
Any unauthorized use of the LASIP routines for industrial or profit-oriented activities is expressively prohibited.



The main routines provided in this DemoBox are the following:

function_AnisSect_explorer new! Sectorial anisotropic neighborhood     Anisotropic NeighborhoodsNeighborhood Explorer
Visualization of anisotropic supports
Shows the anisotropic neighborhoods (i.e. union of the supports of the adaptive-scale kernels) which are used for estimation in the Anisotropic LPA-ICI algorithms.

demo_CreateLPAKernels
utility_DrawLPAKernels
LPA KernelsLPA Kernels
LPA kernel design
Creates LPA kernels and draws them.

demo_DenoisingGaussian DenoisingDenoising demo
Anisotropic LPA-ICI denoising
Performs the Anisotropic LPA-ICI denoising on observations which are contaminated by additive Gaussian white noise.

demo_RecursiveDenoisingGaussian Recursive Anisotropic LPA-ICIrecursive denoising demo
Recursive Anisotropic LPA-ICI denoising
Performs the recursive Anisotropic LPA-ICI denoising on observations which are contaminated by additive Gaussian white noise.

demo_DeblurringGaussian Deblurringdeblurring demo
Anisotropic LPA-ICI deconvolution
Performs deblurring (deconvolution) from observations which are blurred and noisy. The RI (Regularized Inverse) and RWI (Regularized Wiener Inverse) Deconvolution Algorithm with Anisotropic LPA-ICI adaptive estimate selection is used.

demo_DenoisingSignDepNoise Recursive algorithmdenoising demo
Recursive Anisotropic LPA-ICI denoising for Signal-Dependent Noise
Performs the recursive Anisotropic LPA-ICI denoising on observations which are contaminated by signal-dependent noise (e.g. Poisson, Film-Grain, Speckle).

demo_DeblurringPoissonian Blurred observation Signal-dependent noisedeblurring demo
Anisotropic LPA-ICI Poissonian Deconvolution
Performs deblurring (deconvolution) from observations which are blurred and noisy. Noise is modeled as a Poisson process.

demo_InverseHalftoning Inverse HalftoningInverse-Halftoning demo
Anisotropic LPA-ICI Inverse-Halftoning
Reconstructs a continuous-tone image from a given error-diffusion halftone image. Inverse-halftoning is performed using the Anisotropic LPA-ICI deconvolution with RI (regularized inverse) and RWI (regularized Wiener inverse) adaptive-scale estimates.

demo_AnisotropicGradient Riemann surface     edgesAnisotropic Gradient demo
Demonstrates the Anisotropic Gradient concept using the Riemann surface example.

demo_CreateMRLPAKernels MR kernel     MR kernelMR kernels
Multiresolution LPA kernel design
Creates and draws multiresolution (MR) LPA two-dimensional kernels.

demo_MR_FilteringGaussian filtered subband       MR variance mapMultiresolution denoising demo
Anisotropic multiresolution (MR) LPA Denoising
Performs the MR anisotropic LPA denoising on observations which are contaminated by additive Gaussian white noise. Multiscale kernels are used for MR signal analysis and thresholding for noise removal.
Any unauthorized use of the LASIP routines for industrial or profit-oriented activities is expressively prohibited. By downloading any of the LASIP files, you implicitly agree to all the terms of the LASIP limited license PDF.


zip fileCLICK HERE TO DOWNLOAD .ZIP PACKAGE
version 1.20 (released 9 May 2016)


Some simulation results obtained using this software



The LASIP Anisotropic Nonparametric Image Restoration DemoBox contains also efficient routines for:
- calculation of objective error criteria: SNR (Signal-to-Noise Ratio),
PSNR (Peak SNR),
ISNR (Improvement in SNR),
MSE (Mean Squared Error),
RMSE (Root of MSE, l-2),
MAE (Mean Absolute Error, l-1),
MAX (Maximum Absolute Difference, l-infinity)
- robust noise variance estimation: Median of Absolute Deviation (MAD) in wavelet domain,
Fast Laplacian-based techniques using mean in global and blockwise mode,
etc.
- error-diffusion halftoning: Floyd-Steinberg,
Jarvis-Judice-Ninke,
user-defined diffusion kernel