Crosstalk Algorithm - Optical system

[btb4198]

I have a camera that take pictures of square boxes that are really close to one another. There is a light that is Illuminating the boxes but when I try and isolate a single but in my picture, I am reading higher values then what I should be. If I remove the other boxes and only do one box, I am getting lower values. So I believe I have cross talking. What is the best Crosstalk Algorithm I can use?

 

[berkeman]

Can you post some representative pictures and show your brightness data? Thanks.

 

[btb4198]

for the black part, I am getting value like 4.67 and they are lowest when I do not have the white or grey boxes. so when the picture is uniform

 

[phyzguy]

It could be crosstalk and it could be scattered light. What type of detector is it, and what is the detector architecture? Is it read out in a single array, or in multiple segments? Can you illuminate the detector with a small spot of light that you can move around? This is a good way to quantify crosstalk

 

[btb4198]

It could be crosstalk and it could be scattered light. What type of detector is it, and what is the detector architecture? Is it read out in a single array, or in multiple segments? Can you illuminate the detector with a small spot of light that you can move around? This is a good way to quantify crosstalk

Please see attachment. I am using a camera that has a CMOS image sensor. What do you mean by " illuminate the detector with a small spot of light" that can move around ? how small of a light ?

 

[btb4198]

When I put my cell phone in front of the camera everything turns white

 

[phyzguy]

I don't think you are dealing with crosstalk. Try pointing the camera at a dark screen and point a laser pointer at the screen. You should see a spot of light in the camera. If it is crosstalk, you will see "ghost" images of the spot of light.

 

[btb4198]

So I am new to Optics, What is the different between crosstalk and scattered light? Do you believe that system is suffering from crosstalk ? Is there a test I can do for that ?

 

[phyzguy]

Crosstalk is electrical coupling where the signal from reading out one pixel gets coupled into another pixel. You typically see a faint copy of one part of the image in a different part of the image. Since it is an electrical effect, the faint copy is an exact copy of the real image. If the real image is in focus, the crosstalk image is in focus, too. A test would be to do what I said. Have a bright spot in one area and the rest of the image dark. If it is crosstalk, you will see one or more copies of the bright spot in other parts of the image. Scattered light is an optical effect, not an electrical effect. It is light from the bright regions entering into the dark regions. In this case, the light scattered into the dark regions will typically be diffuse and out of focus. Look at the attached images. On the left, there is both crosstalk and scattered light. Notice how the ghosts of the numbers "2" and"5" are still in good focus. On the right, the crosstalk has been subtracted, but the scattered light remains, where there are fuzzy, out-of-focus copies of the bright regions in the dark areas. From what you've showed, I think your problem is scattered light, not crosstalk, but I don't know for sure.

 

[btb4198]

Is there an Algorithm, I could use that could digital remove scattered light?

another question, in my picture, I never see any fuzzy, out-of-focus copies . I never seen any copies at all . but I still get lower values with uniform image vs one with the boxes like I show before. so I know it is there. why is that ?

 

[phyzguy]

It depends on what the scattered light looks like. In the images I posted, those fuzzy out of focus images are actually reflections in the optical system, so they are not completely out of focus. It's also possible to have more diffuse scattered light, which you won't see easily. To subtract it off, you need to characterize what it looks like. For example, suppose you illuminate one part of the image, say white on the left and black on the right. Then, what do the counts in the dark region look like? Are they uniformly high? How much higher are they than a completely dark image? What if the bright region is 25% of the image vs 50% of the image? Does that reduce the counts in the dark region? Do the counts fall off as you move from the edge of the bright region to far away from the bright region? Once you can answer questions like these, you can build a model of the unwanted light and subtract it off.

 

[btb4198]

great! this has been a big help!

 

[btb4198]

Is that an Optical way to reduce scattered light ?