Detecting noise presense on medical images

0
votes

Noise can be different and it requires different techniques to remove it:

  • gaussian noise
  • speckle noise
  • salt noise
  • pepper noise

I've faced with noise samples which look like smth mean between gaussian and speckle noise. It more looks like uniform speckle noise which evenly fills all the space on the image.

enter image description here

What Image Processing Toolbox function or general algorithm can be used to only detect the noise (not remove)? So that it can be measured on the piece of image.

enter image description here

I guess that it would be smth like pixel brightness density (how many bright pixels are presented on the image area) or brightness rarefaction (how much is average distance between bright pixels on the image part). But I think there is more robust way.

An example below is non-noise representation in scale.

enter image description here

1
matlabimage-processingnoise
You'd probably have better luck asking this at dsp.stackexchange.com. However, I'd say that in general noise estimation requires an underlying noise model and signal model. If you don't know the characteristics of either, then it may difficult to make progress. – Oliver Charlesworth

1 Answers

0
votes

As a hacky thing to try. First erode/dilate your image. Then subtract this processed image from the original image. Then use a sliding window and calculate a distribution of of intensities. This distribution might be a good proxy for the noise