Find lint on a display

Hi, I have images from a lc display and I want to find lint or dust or similar defects.

The problem is the quality of the images. I have some kind of gradient in the brightness. Some parts are darker, other parts are brighter and the overall brightness varies from picture to picture. There is also a pattern from the pixel segments.

Here are some examples

This image is good

How can I find this defects? I tried different algorithms like laplacian, but I haven’t found a good solution.

Any suggestions?

Sorry, I only can upload one image per post

This has lint in the upper left corner

There is something on the left and the right side is darker

This one is good, but much brighter

And there is something obvious on the bottom right

first, there seems to be a red pixel stuck near the center of the picture, in all pictures. I’m guessing you need that ignored. that should be fixable by picking that pixel once, and then erasing it with an average of the surrounding ones.

I see some “saturation” effect near the backlight. green saturates already and then red does too. saturation is somewhat undesirable. perhaps tweak the camera parameters so those areas are still in range.

the grids you might be able to attenuate with a fourier transform and careful erasing (nulling) of coefficients that correspond to the grid. or just ignore. or blur (median blur maybe) but that could attenuate the lint as well.

examples: How to Remove Grid Lines from an Image - #7 by pedgalher - Image Analysis - Image.sc Forum

a difference of gaussians (bandpass/highpass) could remove the uneven lighting.

Yes, it is hot pixel in the camera. This can be ignored. Sadly I don’t have access to the camera until now. Maybe in the future.

I can’t use blur, this removes also the artefacts I want to detect. At least the smaller ones.

What do you mean with “erasing of coefficient that correspond to the grid”?

Thank you very much for your explanation. It has at least helped me to use these functions in Gimp.

I think with the current image quality it is extremely difficult to impossible to find these defects.

I was able to manually make the defects visible in Gimp, but automating this with OpenCV is almost impossible.

I was able to manually make the defects visible in Gimp, but automating this with OpenCV is almost impossible.

I think solving your problem is not a Hard one. Could you share your trial codes and results.

why are all the pixels ON when you are looking for lint? turn them off. then there’s no grid.

I don’t have any code. I just play around with Gimp to understand how the filters work.

Sadly the pixels are OFF. The contrast setting is much too high. I know that, but that’s the settings our “experts” are using.

I hope that I can change that in the near future. I see there is room for improvement, but I have to learn more about image processing. That’s why I try to get the best out of the current images.

Btw: I have some good results using a combination ob Sobel X and Y (in Gimp G’mic). This removes the grid and the gradient. But there are still too many artefacts especially at the bottom side. I will upload some examples later.

This is done in Gimp using Sobel X and Y form the G’MIC Convolve filter with Neumann boundary. Doesn’t look that bad.

I tried different combinations of filter, but it is still difficult to find defects.

1 Like