Hi all again.
I discovered that my Unibrain B/W camera only captures at 640x480. This explains why I’ve been having performance issues with the undistort and colorTracker nodes. Even when I change it with the VideoIn pins it hasn’t changed and I just assumed it had. I’ve checked the specs of the cam and it only works in this res.
Is there a freeframe plugin I can use to half it?
I think the GPU method of VideoTexture > scale > AsVideo will slow things down more that what I imagine linking up and extra freeframe will do.
I see the color version of the unibrain will go to 320x240 and 160x120. However, I understand the color version is less sensitive to IR light thought. Does anyone know how less sensitive it is in real terms?
Any advice would be greatly appreciated, thanks.
hm…no, the freeframes cannot resize video. we’d need to have a native directshow resize filter. anybody?
Indeed, I never recognized it before…
Thats not very convenient. But I cant give you comparisions to color. Sorry.
Edit: Did you ask Unibrain Support about this?
frank, I have just emailed unibrain support. I post on here there response. Good idea, I suspect they’ll say it can’t be done. Their RAW color cameras only support 640 too so I think it is all hardware side.
been researching… the CCD sensors in the color only does 400 lines compared to 480 lines in the monochrome. The way the color of each pixel is worked out it also means images are not quite as sharp.
Joreg, I’ve been looking around for a filter but it doesn’t seem straight forward, it’s not an area I worked in before. Also, I see you’ve recently posted “directshow-filters needs to be registered manually in vvvv. you’re welcome to send us your directshow-filters but then they would only work with the next release.”
Does this mean I couldn’t get my 640 resized 320 without sending the filter to you and getting it put in the next release? or am I miss understanding the whole thing?
*I’ve been looking around for a filter but it doesn’t seem straight forward
coding one on your own is really not trivial. at least not for all the different media formats. here they have one for RGB24. if you find out about their licensing and if you test it in graphedit or probably better graphstudio, we could think about integrating that in vvvv.
as integrating a directshowfilter needs some manual coding you’d always have to wait at least for the next beta.
joreg, thanks for looking further into this.
I’ve managed to get the filter you’ve found working inside of graphedit not graphstudio. Graphstudio wouldn’t allow me to add any filters, not even the capture source. Not sure if it is anything to do with XP x64.
I also had to use another filter that converts from Y800(B/W unibrain capture format) to RGB24.
How do I now go from this graph and use it in VVVV?
Also their licence seems to allow you to use it, I’ve attached it if you want to look.
license.txt (1.6 kB)
ok. bsd license is fine. we can use and ship this with vvvv. i will have a look at it for next beta.
that’s cool. I assume that means I can’t go from this graph into vvvv now, I have to wait?
is there a timeline for the new release, I’m also very interested in this filter … at the moment I’m using 2 cams to avoid this problem but that’s not the best way. I will do a installation in one month where it would be nice to be able to use the scale down approach.
Got a reply from uniBrain…
You can see the difference of the CCD sensitivity, at page 7 of the CCD’s specification:
(Please notice that the color RAW model uses the same CCD with color cameras)
The color CCD is sensitive only for red light near IR area.
Unfortunately you cannot get 320x240 frames from the B/W model.
I looked at these graphs linked above and asked…
“From the graphs it looks to me that the color camera is more sensitive to red further into the red that the B/W. So am I correct in assuming that an IR light source is mainly made up from red light, and does this mean I would expect to get more detection of IR from the color camera?”
I am not sure about this assumption. Based on customer’s feedback, I know that the B/W model is more preferable than the colour model.
Additionally, Sony’s specification limits the graph to the 700nm, so I am not sure if we can safely assume even that the red color has better relative response than the monochrome CCD.
Although the colour camera is sensitive to NIR light, the monochrome model is considered to be the best choice for such projects.
I knew this to be the case from somewhere on the net, but I thought I’d ask a little more about it so I could get a better picture of what is going on.
Does anyone else have an interpretation of the graphs?
Is my question to unibrain based on any logic?
Im not sure if you can say that easily color is more sensitive, because its a relative measurement (y-axis). I couldnt find any kind of value or unit.
So 0,62 (color) to 0,52 (BW) means nothing more thna 0,62 and 0,52.
Usually, sensitivity is being measured in Lux, which is an absolute value.
At all, single chip BW cam is always more sensitive than a Color cam, because the use of a bayer pattern in front of the CCD to get colors. Thus you will always get reduced light intensity compared to BW single chip cam, which doesnt use a filter.
good points, well made. thanks.
Didn’t think about the units on the y-axis. I assumed them to be the same.
Yes the bayer pattern will certainly make a difference.
My color version of the cam has just turned up so I will get testing and hopefully my IR source will be bright enough to get picked up. Gutted they don’t do 320x240 monochrome.
If you are German take a look at this site, there it explained in detail how a CCD works and why BW or RAW mode is a better choice than (pseudo)-color (besides missing rescaling options ;)
My German skills are about age 3 I think. I’m Welsh and I can’t even speak/read Welsh. But then who can?
I’ve been testing my new color unibrain camera with my tracking project. It has no problem picking up my IR source. However, it is a strong IR source and I’m only 3m from it but throught some thick spandex/lycra material so I’d expect not to have much problem.
Maybe if I get some time I can do a side by side comparison on them both. It must also depend a lot on what type of visible light filter is used and the frequency of the IR source and other IR light pollution. So it very different for every project.