Could you be more clear regarding what kind of noise are you looking for? And which kind of images are on your interest?

Dear Muthukumaran Malarvel

Most of the equipment's take the images add the different forms of noise and to quantify this noise,  the noise is added to the image (Gaussian with  mean and variance) then the denoising filters are tested for removing these forms of noise.

Good Luck.

Prof. S. EL-Rabaie

You could apply a fourier transformation and check the higher frequency amplitudes for one of the characteristic curves of the different 'colors' of stochastic noise.

Though this method is not perfect, typical images carry the useful content mostly in the lower ferquency area, thus allowing such crude assessments in the higher bands. But be aware that some frequencies may anyway components attributed to the 'useful content' - eg. from edges. Thus I'd try no simple comparison but normalized cross correlation between the spectrums.

Hope this helps

you can try a more simple approach of  analyzing the histogram of the output image. If the noise is of Gaussian type, then the histograms are likely to look similar to Gaussian probability distribution function. also same like distributions like salt and pepper, gamma.

you can use the metrics like Peak Signal to Noise Ration (PSNR) and RMSE for measuring the noise. all you need is the original image and the noisy image. simple. 

the MATLAB program for PSNR and RMSE are directly available in the internet. 

thank you.

Thanks for your valuable replies.

Dear Tilo Strutz,

Yes, I need this. How it can be estimate or identify If a single image contains noise.

Dear All,

Is there any procedure or method to solve this problem? If it is not available at present (in image processing domain) please confirm. 

Thanks & regards

Dear Muthu 

It seems a nice question. I have attached some resources to solve your problem. Go through those. 

Regards

Ehtesham

http://people.csail.mit.edu/celiu/denoise/estnoise/

http://ieeexplore.ieee.org/document/1632204/

Hi there. I guess answers above are more detailed or specific. Though if it was my task I guess firstly I'd try to check if gaussian blur subtraction do something.

To do it you can create two images from you with added gaussian blur (different radius) and substract it from each other. In imageJ you can find it in "process - filters and image calculator". If your image will become drastically less noisy I guess it will indicate it has considerable ammount of blur itself.

Maybe it is to simple, on the other hand maybe its simplicity will suit you.

Best,

V