What Would You Do To Improve This Image?
In running a test with Arecont 5MP and 10MP cameras the past couple of days, we've been running into severe issues with how they handle light (~700 lux on target, indoors). Below is an image with default settings:
Note that even in auto-iris Arecont models, iris control defaults to OFF. So the first thing we did was turn it on, which improved things, but not drastically:
We found that adjusting brightness and gamma had the most effect, but this will obviously affect low light performance:
So, the question is: what else should be done to compensate for this? Have you had similar issues and what did you do to remedy them?
Here's an image of the full FOV (~3' wide), also:
Using a wider, ~9' field of view, the entire target is washed out. We used the images above simply because at wider FOVs smaller text is difficult to read regardless of light handling issues, due to the lower PPF.
Also, just to pre-emptively head off any sarcasm: don't say "use a camera other than Arecont."
I have nothing to add to the discussion, then....
Ethan, could you please make another picture with also the other side of credit card?.. It could be highly interesting, maybe not in order to improve the luminosity, but.....;))
What is actually the final goal here? To reach the best real-image quality possible, or just to read and recognize better the items?
Specifically, the point of this test was to show legibility of varying text sizes at different FOVs and resolutions. Not to show light handling issues with specific cameras.
So I guess the answer is the goal is to make that text as legible as possible.
Is the illumination around a constrain? Or better, are there limitations in illuminating the items in a certain way or in another?
My experience (and you teach it much better than me..) is that anyway clear_contrast_visibility_in_low_light / high_resolution is always a trade-off with any brand of videosurveillance camera (thus, not talking here of course about specific cameras for industrial vision)..
Thus, more than the cameras parameters that I suppose will have anyway a limit and trade-offs, the goal according to me is better reacheable here surely thinking also and above all about how to increase the "collaborativeness" of the environment. So for example the quality of the illumination (quality, not only quantity: clean white deep lights, leds, ..). If possible, of course..
Off the top of my head, I'd say the overall exposure of the scene is your main enemy - most of it is dark grey to almost black and that's undoubtably wreaking havok with the camera's exposure modes.
So first thing I'd do, if you want to maintain the FOV and overall contrast, is change the camera's metering mode, if possible - go to a spot or center-weighted mode, or define the text area as your exposure target...
If that's not an option, and all you have is full-frame averaging, the next step would be zooming in more on the lighter objects to reduce the black area in the frame.
Alternately, you change the background to a lighter shade of grey so the camera isn't overexposing to compensate.
Also, see if the camera has adjustable gain limits, which may allow you to reduce the maximum gain while leaving low-light performance unaffected.
AGC, AGC, AGC. I've said it before and I'll say it again - if the camera has AGC controls, turn them down or "Off".
What???? What does this scene have anything to do with gain control? I doubt gain is being applied as this scene is overexposed to begin with.
Btw, I think Matt's suggestion of an exposure zone makes sense though it's obviously going to be at the expense of the rest of the image / scene.
To my knowledge there are no exposure zone or metering settings in the web interface, but according to this video, you can set an ROI via software. I'll check it out tomorrow.
If you were Arecont Vision certified, you would know this....
Lots of cameras have this and I agree it will make a difference but in this case, the conclusion is fairly obvious, the rest of the scene is going to get very dark. Also what happens when lighting levels change throughout the day?
If you really need to properly expose the text AND the surrounding black card in this scene, you're probably looking at a WDR camera to start with. For the sake of this testing, I don't think the clarity of the black base is a concern, is it?
The black base is not a concern if all we want is the 1 foot by 1 foot piece of paper. Surely, though, it is a concern for real users who will want to cover a larger area.
Btw, anyone care to guess why the Avigilon 5MP and 10MP cameras are having so much problems here but the HD cameras in our test did not. Matt's voting for WDR? Any other suggestions?
Right, but the point is, how many situations are really going to need to properly expose BOTH of these VERY contrasty areas? Even in the standard WDR situation - backlit people at building entry points - we're only really concerned in properly exposing the faces, not the background.
And anyway, this question just has to do with Ethan's issues with this test setup, where all he's concerned about IS the one-by-one-foot piece of paper.
"And anyway, this question just has to do with Ethan's issues with this test setup, where all he's concerned about IS the one-by-one-foot piece of paper."
Disagree. We cannot and do not do tests in a vacuum. What is the point of us doing a test that sharply misrepresents real world conditions i.e., monitor a 'one-by-one-foot piece of paper'?
a test that sharply misrepresents real world conditions
Whoa Nellie! I figured this test was on account of that fella's question couple days back. Either way they seem as close as kissing cousins.
I am taking classes and trying to participate. Seemingly F-stop would play a role in this discussion as relates to image quality or not? Would or could you change the lens to improve quality?
Do we ever know what chip sets come in all these different camera manufactures? If we say they are equal then evidently cameras must be designed better than others behind the chip set consequently not created equal. I am off base? Comments welcomed please.
You doubt, but do you know it for a fact?
It is, Jim!
This discussion is simply because we hit into the issue that Arecont 5MP and 10MP have glare / imaging issues in the scene.
We'll be posting the full test results in the next week.
Good point. If we used a manual iris, we could close down the iris / increase the F stop. That would likely reduce the overexposure on the piece of paper. Of course, it would permanently make the scene overall darker which would likely be a very bad thing if the area ever had less light.
It's very hard to know what chip set comes in different cameras. Some manufacturers, especially those without brand names, will brag about their chip supplier ("It's a Sony Effio"). However, usually prominent manufacturers will hide that, focusing their marketing on their own brand.
To your point about chip set, I do suspect that there is something about the 5MP and 10MP imagers that they are using that is causing this, because the HD cameras we are using do not have this problem.
As expected exposure reference does help in viewing the text substantially:
The big tradeoff, though, is that everything else is much much darker, seen in this 3' FOV shot:
For reference, here is the same scene with no exposure region set:
I give up.
I answered Ethan's question as he posed it.
Now you're changing the question and telling me I'm wrong.
Ethan, if you'd like any more suggestions, feel free to email me. I'm done with this thread.
Your suggestion is appreciated. I am just trying to explain the full context of what we are trying to do in this scenario. The picture Ethan shared is part of a larger test scene.
Started by John Honovich
|less than a minute by Undisclosed #4|
Started by John Honovich
|about 2 hours by Undisclosed Manufacturer #10|
Started by John Honovich
|43 minutes by Jarad Regan|
Started by Carl Stoffers
|about 2 hours by John Honovich|
Started by Undisclosed Distributor #1
|less than a minute by John Honovich|
Back to Top