Adding poisson noise to image data store

Is possible to add poisson noise to the imageDataStore to generate more data for training a network?

Answer 1

[MathWorks Support Team]

This is possible using the transform/combine methods that were added to Datastore in 2019a, together with this and the "imnoise" function in the image processing toolbox can be used to add Poisson noise to an image to simulate that noise model for denoising workflows.

Below is an example, which walks through on how to use transform and combine to implement a data pre-processing pipeline for denoise problems with an arbitrary noise model:

https://www.mathworks.com/help/deeplearning/ug/image-to-image-regression-using-deep-learning.html

And in the “addNoise” function, use the ‘poisson’ option in the IPT function "imnoise" instead of ‘salt & pepper’:

function dataOut = addNoise(data)
  dataOut = data;
    for idx = 1:size(data,1)
      dataOut{idx} = imnoise(data{idx},'poisson');
    end
end