5

I understand how to generate noise or filter out noise on an image, but if I have two images at difference noise level, for instance one photo taken during daytime and the other taken at night, how can I tell which image is more noisy in OpenCV or Matlab?

Thanks!

Kai Chen
  • 51
  • 1
  • 2
  • 1
    Quite broad, but you can start with: [How to calculate an image has noise and Geometric distortion or not](http://stackoverflow.com/q/14808033/5008845), [Noise Estimation / Noise Measurement in Image](http://stackoverflow.com/q/2440504/5008845), and on Matlab FileExchange, [Noise Level Estimation from a Single Image](http://www.mathworks.com/matlabcentral/fileexchange/36921-noise-level-estimation-from-a-single-image), [Fast Noise Estimation in Images](http://www.mathworks.com/matlabcentral/fileexchange/36941-fast-noise-estimation-in-images). Plus a ton of papers, google for that ;D – Miki Feb 25 '16 at 23:18
  • 1
    A simple tools would be to compute Average(|Image - Median(Image)|). So you apply a denoising filter (median) and you compute the difference with the original image. It's simple and fast, and it gives you an idea of the noise. – FiReTiTi Feb 26 '16 at 00:12
  • shoot for each scenario 2 (or n) images with fixed camera position, fixed object scene and fixed illumination. those pictures should be identical if there is no noise, so every change in intensity/color is noise instead. This doesnt work if you only have a single image per scenario. – Micka Feb 26 '16 at 07:05

1 Answers1

2

You can compute signal to noise ratio in flat regions(with no edges) of an image. Compute mean and variance of the signal in flat region. Compare similar flat regions. If an image has higher SNR than other image, it means that it is less noisy. To visualize this you can crop flat region in both the images and check their histogram. A noise-free image should have a narrow gaussain distribution ( low standard deviation) in flat regions compared to a noisy image which will have wider gaussian distribution.

igweyn
  • 141
  • 5