Programmatically generating over/under exposed pictures

Is there a simple way to introduce realistic over-exposure / under-exposure distortion in images? That is, synthetically over/under expose images such that the resulting distorted pictures are very close to what a camera would generate.

I could think of shifting the mean of the entire image in all three color channels. For example: if we consider a test image (I),

OE = I+75;
UE = I-50;

OE will be globally much brighter with certain regions white-washed. But this is a global operator. I am wondering if there is a way to locally introduce these distortions in the images, such that the distortion appears realistic. Is there something in MATLAB / Python that achieves this?

Answer 1

[Kushagr Gupta]

There are various ways in which local exposure distortions can be introduced to an image.

One of the ways would be:

1. Create a binary mask of the same size as the original image.
2. Randomly choose size and location of patches (rectangular are the easiest, would just need to select the indexes in a matrix)
3. Assign some value (say 75 for over exposed) and add the binary mask to the original image.
4. This can be done in a loop and the randomness would help in creating different distortions for each image if need be.

In terms of code, taking an example, it would look like:

I=(imread('coins.png')) 
Imask = zeros(size(I));
Imask = uint8(zeros(size(I)));
Imask(20:40,60:90)=75;
imshow(I+Imask)

This is just an illustrative example and should serve as a starting point.