I have some data from thermographic analysis, stored in a third dimensional tensor (pixel-pixel-time). The tensor contains just the amplitude of the signal captured by the thermocamera during the experiment.
In post-processing I am using the following steps in the following order:
- I am removing the background noise
- For each signal contained at each pixel (512 x 640 pixels) I am applying a moving average to smooth the data
- I am removing the gaussian noise
- I am removing the salt and pepper noise
- I am applying a custom made LowPassFilter at very low frequency
(eventual comments and feedbacks on the order are welcome)
Now, to my understanding, in order to remove the gaussian noise (GN), one should know the size of the sigma.
But to my knowledge, I have found this way to remove the GN but I do not undestand how the sigma is eventually taken into account.
for i = 3000
denoisedImages(:,:,i) = conv2(double(noisyImage(:,:,i)), ones(3)/9, 'same');
Is this the right way to remove the GN?