Remove background diffraction patterns from image

Hi, we are doing a project of imaging things using laser as a light source and with an aparture to help "remove all to big diffraction patterns" in the image. When images is obtained, we would want to remove the unwanted diffraction patterns in the background of the images. We have inserted 2 images here, the one named "JustBackground" contains the diffraction patterns (where just the laser hits the "detector") that we would want to remove from other images such as "IMG_6610". We have tried subtracting, filtering and imabsdiff the two images and have failed.

Could be that we have used these methods wrong in some way. Please help if someone has any suggestions as how to remove the "background" in images.

Thank you in advance!

Sincerely

Answer 1

[Image Analyst]

Have you tried dividing the images?

[Sasithon Meesan]

Thank you for your suggestion! We have now tried your suggestion, however, the images are not of the colors that we would want them to be. We would just want to remove the diffraction patterns but keep the same red and black colors in the images.

OriginalImage = imread('IMG_6610.jpeg');
Background = imread('JustBackground.jpeg');

%Divided_image = (double(OriginalImage)) ./ (double(Background));
Divided_image = OriginalImage ./ Background;

subplot(1, 2, 1)
imshow(OriginalImage)
title('Original')
subplot(1, 2, 2)
imshow(Divided_image)
title('After')

[DGM]

In order for any method to work with a reference image, the images need to be registered. If this is taken off-axis for reasons of physical interference with the projection, that's fine, but the images need to be aligned.

The best start is to have the shots taken with the camera in a fixed position. Alternatively, they can be aligned and the perspective corrected afterwards, but there needs to be some reference in the image to allow them to be aligned if they aren't already aligned. The obvious feature to use would be the corners of the projected image, but they're cut off.

Can we attempt more elaborate methods of registration estimation? Sure, but you're creating another entire task obligation that shouldn't even be necessary.