Mirage Recorder

[Take Vos]

When a dark image is shown my bad pixel detector basically disables all pixels and thus each pixel will be interpolated from its neighbour pixels, basically dropping in resolution.

This is not really a problem as it removes a lot of fixed pattern noise and smoothes out the image. But it will also show these blobs of noiseless dark patches which are a bit distracted compared to the rest of the image.

Luckily I already have measured the amount of white noise during calibration, so the only thing to do was to add a random amount of that white noise back onto the interpolated pixel. Now the patches are gone and it looks a bit more uniform.

Cheers,
Take

[Paul Curtis]

Take,

I hope im not polluting the thread (well i am) but this lens looks like it might be a good lens, the charts look pretty good and on a 1" should be better.

http://stilar.de/hp32638/Stilar_-2_8...bd6e8392XY25be

No idea how much though...

cheers
paul

[Take Vos]

Hi Paul,

This is not really pollution. I like to have a lively discussion on my thread.
Anyway, interesting lens, to bad there is only a single lens and not a whole series available, it is also very wide.

I think I've seen 1.2" lenses before, which could be interesting if I am ever going to the 2048 pixels wide sensor. But of course the alias filter for a 1.2" lens is probably wrong.

Cheers,
Take

[Paul Curtis]

Take

http://docter-optics.de/hp834/TEVIDON_-CCD-lenses-.htm

from the same people, so there are some others (of which only a couple apply to a 1")

When you say alias filter do you mean a OLPF? Does the Pike have something like that already? Does it have an IR filter? (you can get filters that do both)

cheers
paul

[Take Vos]

Paul,

Yes, I do mean OLPF when I say alias filter.

The IR filter for the pike is glued (or something) to the C-mount. The C-mount itself is mounted into a wider screw mount, with very narrow threads for back focus.

I do not see a way to add or change the build-in IR filter, but I may not have looked at it closely enough.

[Paul Curtis]

Take

the OLPF should relate to the sensor itself rather than the lens, it sounds like the Pike (or in fact any of the machine vision heads) do not include one.

This may cause aliasing with camera movement (but perhaps the debayering process helps here).

You can get combined IR and OLPF, so if it was possible to remove the one there then you could if needed swap it out. Or perhaps that IR filter is in fact an OLPF as well.

Is anyone aware of a problem with aliasing and these bayer cameras?

cheers
paul

[Jason Rodriguez]

Yes . . . you need an OPLF if you want to avoid aliasing artifacts, especially chromatic aliasing, and the dreaded moire (which couples into the luma channel as well, and is impossible to get rid of through better debayer algorithms). So in other words, on objects that produce the really bad chroma moire, the chroma artifacts can be compensated for in the debayer algorithm, but the luma artifacts left behind can't.

You can get combination OLFP/IR-cut filters (this is what we use in the SI-camera most other camera manufacturer's do this as well). Also typically the OLFP is made for the pixel-pitch of the sensor (in the case of the Kodak CCD, that would be a 7.4um pixel-pitch). Beware though that the inclusion of the OLFP creates a refraction index, so the length of the back focus gets a little longer and must be compensated for (i.e., you can't keep the same back-focus distance that you had with no filter).

Thanks,

Jason

[Take Vos]

At what distance from the sensor should such a OLPF filter be placed?

[Jason Rodriguez]

As close as possible without creating unnwated optical artifacts (typically flaring) . . . . typically it's a couple millimeters away from the sensor . . . too far away, and you loose the effect of what the OLPF was suppose to-do, which is to basically make the light that would have gone to just a single photosite, and spread it over the surrounding 3x3 photosites (to prevent aliasing). To far away, and that light will not be effectively targeting the surrounding 3x3. This is primarily a Nyquist cut-off filter, not a "blurring" filter, i.e., the idea is not to "blur" the image, but to prevent high frequency detail that is beyond the Nyquist resolution limit of the sensor from being detected by only a single photosite (which produces aliasing) . . . any given piece of detail that the imager can see should be slightly filtered to the surrounding 3x3 pixels.

If you think about it, a bad or slightly out-of-focus lens also serves the purpose of a OLPF, and if the pixel pitch is smaller than the resolving power of the lens, the same thing occurs. But ideally an OPLF does not make the image "blurry" like these examples I just listed.

In fact, a proper OLFP should accentuate detail that is below the Nyquist resolution limit of the lens by preventing false aliasing artifacts, moire, etc, from "reflecting back" and corrupting the "real" detail that the lens is resolving.

Thanks,

Jason

[Take Vos]

I've been working on a new implementation of a debayer, I now have an very standard version of AHD implemented using the Acceleration framework, which doubles the speed compared to my old algorithm. The Acceleration framework consists of routines that make simple arithmetic operations on the whole image at once, such as subtracting one color plane from an other color plane.

Still, the refinement step of AHD is quite expensive and difficult to implement using the Acceleration framework.

One change to AHD is that I have added a cross interpolator, that gets rid of most of the maze pattern in images. I could also easily add diagonal interpolation if the needs arrises.

I am still waiting for the anton bauer battery, to see if the noise will be reduced by this.

Cheers,
Take

[Paul Curtis]

Take,

Glad to hear of some progress, do you have any more samples to see yet?

Do you think the noise is power based then? Are you running off mains supply at the moment?

Isn't temperature a big issue? Wrap some cold ice packs around the head to see if you get better noise?

cheers
paul

[Take Vos]

Hi Paul,

I think most of the noise is power or ground loop based, it is just a feel I have when I look at the moving picture. I have an electronics engineering degree, so I do have a little bit of feel for it, of course I could be completely wrong.

The noise is also there when I plug in the camera when it is still cold. So it seems the noise from the sensor is not that much yet compared to other noise.

I have tried running the camera + notebook from the notebook battery alone, without external power. It doesn't help anything. I did notice changes in noise when wiggling the firewire cable.

So my next step will be a complete electrical separation, powering the camera from an battery and using a fiber cable for firewire.

I like the low-tech solution of ice packs. I wonder how long it takes for one of these packs to warm up from the camera. Powering an peltier element from a battery is not ideal.

[Jason Rodriguez]

BTW, Take, depending on how they generate power on the camera, it could possibly be the power-supply itself on the camera (i.e., a noisier switching power-supply rather than a linear regulator generating the analog 5V supply for the CCD, and not enough isolation to keep a clean 5V rail, or possibly noise from the ground of a switching power-supply that is acting like a RF generator and coupling into the ground of the 5V analog supply for the CCD because there is not enough isolation there). Since CCD's typically need a couple different voltages for bias currents, ADC's, clocks, etc., rather than the single voltage supply to a CMOS, they may have opted for regulating the power to those devices using something that might not be as electrically clean as it could be.

This is all a guess BTW, I don't know how they are regulating power on the camera . . . and I would assume that Allied has done the best they can to supress noise issues like that.

Have you talked with them about the noise issues? Can they send you any image samples from that same camera model using your image window and clock-settings without the noise patterns? That would immediately point to an issue with your setup or camera head itself vs. an issue with the camera model in general.

Thanks,

Jason

[Take Vos]

Hello Jason,

They have told me about a ground loop issue with the firewire cable. And that it would be best to ground both the computer and camera head chassis together.

I also thought it would be silly to make 5V out of a 12-18V power input using a switching power supply. But then I remembered that firewire delivers 30V, then a switching power supply seems less silly all of a sudden.

Maybe powering it from batteries it will use a power regulator instead.

Thank you for you suggestion of asking for an example image. When I still have these noise issues with battery power I will ask for it.

Cheers,
Take

[Take Vos]

I thought I would share two screenshots of Final Cut Pro.
The first screenshot is from the real-time rendering, for which I try to get the colors to match with the second screenshot which is carefully rendered.

Also someone asked for how a picture would look straight from the camera without doing multi-stage white fielding. As that would be an expensive operation, the first image is pretty much straight from the camera (after subtracting black lines)

http://www.xs4all.nl/~takev/fast.png
http://www.xs4all.nl/~takev/slow.png

I still have a miss match between the two pictures which I am unable to find yet, it has probably something to do with how the format the YUV format is encoded. It is a little bit sad that they haven't documented this pixel format anywhere.