Noise estimator / estimation for various types of noise

Code of a new noise estimator and code of some other noise estimators, comparisons.

NOISE ESTIMATION distribution

This a companion distribution of the Ref. paper:

O. Laligant, F. Truchetet, E. Fauvet, 'Noise estimation from digital step-model signal', IEEE Trans. Image Processing, 2013 Dec., 22(12):5158:67

Contact : [e-mail]

This distribution permits to:

- introduce a new noise estimator (NOLSE) with interesting performances on various types of noise

- test various noise estimators on real images corrupted by various synthetic noises

- estimate noise level in image with various noise estimators

The results can be used for various applications. The title image shows an example of image restoration where the parameter of the restoration method is obtained through the noise estimators.

Estimators:

- nolse.m, fnolse.m like-script and function versions of the new estimator NOLSE

- averageN.m noise estimation by S.I. Olsen (see help of averageN)

- FNVE.m noise estimation by J. Immerkær (see help of FNVE)

- mad.m noise estimation by D. L. Donoho (see help of mad)

- TaiYang.m noise estimation by S.C. Tai and S.M. Yang (see help of TaiYang) ;

Rk : please see the above Ref. for the correct equation of the Tai-Yang estimator.

Main:

- testAllGWN.m : test of the estimators on an image corrupted by synthetic GWN

- testAllSpeckle.m : test of the estimators on an image corrupted by speckle noise

- testAllPoisson.m : test of the estimators on an image corrupted by Poisson noise

- testAllImpulse.m : test of the estimators on an image corrupted by impulse noise

- estimNoise.m : estimation of the noise level in an image with various estimators

Tools:

- binarise.m provides a binary image

- fit1p2d.m polynomial fitting

- histo.m histogram (variant)

- jordanOL.m jordan resolution (variant)

- mse.m mean square deviation calculus

- thresh.m low thresholding

Test images are included in the distribution