2
$\begingroup$

I am trying to image some ChArUco targets using a monochrome machine vision camera with a CMOS sensor and the attached images show the artifact I am observing:

2-pixel-wide bright vertical streaks adjacent to light-to-dark edges (only on light-dark transitions from left to right).

2-pixel-wide bright vertical streaks adjacent to light-to-dark edges (only on light-dark transitions from left to right)

Here is another example where the white-to-black edge is nearly vertical, and the contrast is higher.

The "ghost shadow" of the edge is less perceptible but still there

The "ghost shadow" of the edge is less perceptible but still there

Any ideas as to the cause and remedy?

I have tried changing the camera's gain, exposure time, lens aperture etc. But, all these settings do is change image exposure and contrast, which can mask the effect, but not eliminate it (as in the 2nd example image).

I have received some guidance suggesting the streaks may be caused by the over-exposure and saturation of the white regions. I will try and reduce the exposure...

$\endgroup$
6
  • $\begingroup$ It’s a 4-pixel-wide pattern. Maybe your sensor has one AD converter per four pixels? The first two pixels converted show less smearing than the other two. But why is it only on the diagonal edges, not the horizontal ones? I still think that it’s caused by overexposure, but I cannot explain exactly what is happening. Do you have more information about the sensor, other than “CMOS with a rolling shutter”? $\endgroup$ Mar 31 at 19:10
  • $\begingroup$ @CrisLuengo First, thanks for your help and interest. I agree, overexposure does seem the most likely culprit, but it will be next week before I can work on this again. The sensor is an On Semi MT9P031 in a Basler Pulse puA2500-14um camera. $\endgroup$
    – sg65
    Mar 31 at 19:24
  • 1
    $\begingroup$ speak to Basler. they sold it to you, they should know what's going on, or have a real interest in solving this for you (and all their other customers) $\endgroup$ Mar 31 at 20:20
  • 1
    $\begingroup$ oh btw, it could be PWM-dimmed light showing up in the row scan of the camera. your computer display could also be such a light source. shut everything off, use daylight. if the effect is gone then, you'll know it's artificial light. the image looks overexposed already, so if there's more light, it could cause this type of bleeding... per row, whenever the light is on while that row is scanned. $\endgroup$ Apr 1 at 14:18
  • 1
    $\begingroup$ @TimWescott - I bought the camera new from a Basler distributor, not Basler themselves... Changing settings back to factory defaults isn't hard and that doesn't help. I think I possibly have a bad unit; might be worth pestering Basler again. $\endgroup$
    – sg65
    May 15 at 12:17

1 Answer 1

2
$\begingroup$

Probably not a definitive answer, but as close to one as I can get after various tests:

  • It is not exposure. Over-exposure/saturation only makes the problem more easily visible. Though hard to see, the artifacts are still present in a not-saturated image.not-saturated image
  • It also isn't room-light flicker. Daylight tests weren't feasible, but I tried exposure times as long as a second (with zero analog gain, and almost fully stopped down aperture). This would average out any flicker/PWM, but the streaks were still present.
  • Basler were unhelpful - they would only talk to me if I purchased the camera from them, which I didn't. The re-seller's support staff also leaned towards over-exposure, but I probably led them that way with my question.
  • I believe it is my hardware, maybe even the one camera I have. I tried a Basler Ace camera, which has the same sensor, and there is no streaking, even in over-exposed images. enter image description here This camera has a Gig-E interface as opposed to USB3, but that should be irrelevant.

I had hoped what I was seeing might be a known "edge" case with CMOS sensors, but I guess not! Thanks Cris and Chris for your suggestions.

$\endgroup$
1
  • $\begingroup$ I'm working with image sensor chips these days, so I'd be interested to know if an image that goes sharply from white to 50% gray to black shows the stripes on the gray-black transition. $\endgroup$
    – TimWescott
    May 12 at 19:11

Your Answer

By clicking “Post Your Answer”, you agree to our terms of service and acknowledge that you have read and understand our privacy policy and code of conduct.

Not the answer you're looking for? Browse other questions tagged or ask your own question.