Filter and measure MSE, PSNR, SNR of single image

clc;        % Clear the command window.
close all;  % Close all figures (except those of imtool.)
clear;      % Erase all existing variables.
workspace;  % Make sure the workspace panel is showing.
fontSize = 22;

%Load single MRI image
I = imread('IM_00042.tif'); 

% addition of graininess (i.e. noise)
I_noise = imnoise(I, 'gaussian', 0.09); 

% the average of 3^2, or 9 values(filters the multidimensional array A with the multidimensional filter h)
 h = ones(3,3) / 3^2; 
 I2 = imfilter(I_noise,h); 

% Measure signal-to-noise ratio
img=I;
img=double(img(:));
ima=max(img(:));
imi=min(img(:));
mse=std(img(:));
snr=20*log10((ima-imi)./mse)

% Measure Peak SNR
[peaksnr, snr] = psnr(I_noise, I);
fprintf('\n The Peak-SNR value is %0.4f', peaksnr);
fprintf('\n The SNR value is %0.4f \n', snr);
fprintf('\n The MSE value is %0.4f \n', mse);

   %Plot original & filtered figure
   subplot(1,2,1), imshow(I_noise), title('Original image') 
   subplot(1,2,2), imshow(I2), title('Filtered image')

    text(size(I,2),size(I,1)+15, ...
      'Gaussian = 0.09', ...
      'FontSize',10,'HorizontalAlignment','right');

[Latha]

Mse is not calculated from two images

[Image Analyst]

In her code it's not, but in general it often/usually is computed from two images, a reference image and a test image. Her code just uses one image, and she makes the common beginner mistake of thinking the standard deviation of gray levels in an image is all due to noise. Obviously it's not, except in the special case where you know for a fact that your scene is totally uniform (or should be if it did not have any noise).

She should use the functions immse() and psnr() in the Image Processing Toolbox, after getting a better understanding of what those quantities mean, and what she needs to measure.

[Syed Zenith Rhyhan]

Thanks for the solution

[Anirudh Balaji]

@Image Analyst I am trying to estimate SNR of a noisy and reconstructed MR image and there are two methods that i find online.

One where I choose a single image, select ROI-1 of a specific dimension inside the object and then choose ROI-2 of same dimension of the backgorund in the image, and then SNR = [Mean(ROI-1 minus ROI-2)/std(ROI2)].

The other method is simply use reconstructed as reference image, use noisy image as the other and use [peaksnr, snr] = psnr(Image, Ref) to estimate the SNR between two images.

could you please give some suggestion on when to use single image method and when to use multi image method? my ultimate goal is to find if the reconstructed image has higher SNR than noisy image.

Thanks in advance

Answer 1

[iza]

Any comment on this?..

[Image Analyst]

Why should we comment on it?

[iza]

Is it correct way to measure MSE,SNR and PSNR?

[ari ali]

Hello

I have a project about ( image enhancement using modified histogram equalization ) i must be found function (snr , psnr and mse ) for original image , for image after addi noise and for image after applying filter .. in this step i need your help how can find functions for orginal image and noise image and image after applying filter ???? . Best regards

[Image Analyst]

There are functions immse() and psnr() in the Image Processing Toolbox.

[ghada sandoub]

are these functions work with the colour image? i.e. can these functions estimate the MSE and PSNR for the whole RGB image or we have to use these functions separately for each channel of the image?

Answer 2

[sam alsalihy]

noise image not effect on the result

why?

Answer 3

[Hasan Deen]

thank you