Most reasonable DSLR for multi-cam face capture setup

[meshmaster]

hi all,
i'm Ruslan's business partner and just wanted to add we're using the Nikon 50mm f/1.8 E pancake lens in that shot he posted.
thanks for sharing info and helping us get our heads around theses D3200's.

Hi Bernie,

Thank you very much for the information, I really appreciate it!

Cheers,

Joe

[Infinite]

hi all,
i'm Ruslan's business partner and just wanted to add we're using the Nikon 50mm f/1.8 E pancake lens in that shot he posted.
thanks for sharing info and helping us get our heads around theses D3200's.

Hi Bernie,

Thank you very much for the information, I really appreciate it!

Cheers,

Joe

Your newest shot looked better Joe. It would be easier to tell compared to that previous girl though.

Here is a direct comparison I found from D3200 and 600D on the same Rig, same lighting, same "settings"

D3200

http://www.ir-ltd.net/uploads/013_D3200.JPG

http://www.ir-ltd.net/uploads/013_D3200_7.JPG

600D

http://www.ir-ltd.net/uploads/01_Camera-401.JPG

The comparisons are subtle but the D3200's are softer, which makes a difference. Subtle but it does during build.

Also notice the noise / grain present 013_D3200_7.JPG in the blacks and that's at ISO100! notice the orange / yellow tinge. White balance was set to Flash at the time.

[meshmaster]

I agree Lee.

Obviously, there is no question that the supplied 18-55 kit lens are rubbish.

I'll try and shoot the girl again as soon as I can with my vintage 50mm Nikkor.

Thanks for the images Lee, they look pretty damn good but I still prefer the Canon's ! 

Cheers,

Joe

[polygon]

Hi Lee?
   I saw a support bar on the back of the neck in your pictures,  Does that mean the shutter speed is not fast enough?

[Infinite]

Hi Lee?
   I saw a support bar on the back of the neck in your pictures,  Does that mean the shutter speed is not fast enough?

No and yes it's to keep the subjects within the focus / capture volume and also to aid in FACS alignment, as I tend to take 45-100x scans per subject for clients, mostly. It depends on the project. Also useful for multi-lighting.

[bisenberger]

A while ago, when I studied resolution charts at dxomark website, high f-stop seemed to be lens equalizer - pro lens would peform infinitely better at f4-8, but light diffraction at f11 and up seemed to degrade image quality to comparable levels with cheap lens.

Can anyone share their experience, or better yet, share comparison pics of cheap and good glass at higher fstop?

I think this is true for full frame and higher MP. But anything below f8 is useless for full body or face scanning as the focus range is so small. Anything blurred will create highly noisy scan output 

Hi Infinite,
I'm curious what focal length you are using that you need the higher F stop for increased DOF. Wide angle lens have good DOF at lower F stops.
Thanks,
Bill

[Infinite]

Hi Bill,

50mm. The focus range is very narrow below f10.

Thanks,
Lee

[bisenberger]

Thanks for the quick response Lee. I visited your web site, amazing work! 

[Infinite]

Thanks for the quick response Lee. I visited your web site, amazing work! 

Thanks. It's only mainly been possible thanks to Agisoft Photoscan amazing software.

[lgs777]

How about sony A57 guys.. I only have this one.

[ikercito]

Hi everyone, I've reading this topic thoroughly the last couple of days, as I?m trying to get onboard with 3D scanning based on photography. Agisoft looks like the best cost effective alternative around.

I'm a photographer myself, and I'm wondering if any of you has taken into consideration sensor sizes in relation to Depth of Field. The bigger the sensor size is (35mm or APS-C common in DSLRs), the shallower the DOF. That's why I'm thinking smaller sized P&S cameras would prevent the problems with DOF posted before, while retaining reasonable apertures to avoid subject motion. It's obviously a balance between DOF and quality... But I'd reckon a good Canon G15 could very well make the cut. Anybody thought of this kind of alternative?

[chadfx]

Hey Ikercito, I've kept that exact aspect in mind (deeper DOF with smaller sensors) and worked with some other alternatives.

I haven't done as much extensive research as many of the others here (especially for full body capture). Overall, to me there seem to be tradeoffs with all approaches. As you note, the bigger sensor cameras require smaller apertures to achieve deeper DOF. I've used an Olympus OMD micro 4/3 camera for a number of objects and it generally has worked fine as well, and gives you a bit more DOF at similar apertures. However, it also suffers from diffraction issues more readily than larger sensor cameras. So I usually can't go smaller than f8 or f11 on the Olympus, where I can do down to f16 or f22 on my Canon 5D2. So I sort of feel that it becomes a bit of a toss up as to which is better for yielding the best DOF results.

I have also used a Canon A3300 IS, which is able to shoot RAW images when used with the CHDK firmware (http://chdk.wikia.com/wiki/Main_Page). It has an even smaller sensor than the G15, but no real aperture control per se. It also can create fairly well detailed objects if photographed decently (see attachment).

There are also noise and dynamic range issues along with overall image quality and resolving power to consider. Overall I can get sharper, less noisy images and hold more shadow and highlight detail in the images from the full frame Canon. The Olympus micro 4/3 is very close in most aspects, though. Dropping all the way down to the Canon P&S, (purchased refurbished directly from Canon for under $60!), it certainly is the lesser camera in all aspects, but it CAN generate good enough images in many cases depending on your needs.

The G15 has a slightly bigger sensor than A3300, so that might help with some of the image quality issues, but not give up too much of the DOF capabilities. You also might look into the Sony RX100, which has a bigger sensor, but still not as large as micro 4/3, etc.

When it comes to full body capture with dozens of cameras, you will likely need to step up to the Canon or Nikon DSLRs as they have the most complete support for remote control and synchronization. I have done facial captures with 12 of the little Canon's with USB triggering and CHDK, but I wouldn't say that it would be good enough to sell as a service, compared to what most of the others are doing out there.

Not sure if any of that is helpful, but it would be great to hear how it goes for you.

Cheers, -Chad

[ikercito]

Thanks a lot for your info Chadfx, that's really valuable. Being a newbie myself, I was just wondering if that option was considered. I see that lens sharpness is one of the key facts to get proper scans, together with a high pixel count. That certainly screams DSLR. There's and important advantage in Dynamic Range with DSLRs compared to P&S, but I guess that could be solved with careful lighting and exposure. Same for noise, as long as the camera has a clean low ISO. (So forget about cheaper ones..)

I've done a few tests with the camera i had lying around (Lumix LX5) and there's certainly a huge load of noise in the scans, probably because i didn't have time to light properly and quickly processed the RAWs. I'm coming to terms with the whole process, it's quite a steep learning curve heavily influenced by the hardware quality i guess...

I was thinking of building a no frills camera array with point and shoots, similar to yours, but for now the LX5s have been ditched (no way to release remotely, nor sync with other cameras). I've never used CHDK, so making an investment in a dozen or so Canon compacts is giving me chills for now. Doing it with DSLRs is clearly out of the way for now. I have no intention to give it as a service, there's still a lot to learn. I'll keep testing.

Thanks everyone!

[tommyboy]

Lee/all, one more question about the Nikons for y'all if you don't mind!  I dig the Canons.  However I've done several tests with 6 each of Canons and Nikons stacked right on each other in stereo pairs, and the Nikons generally result in 20-30% more points resolved, and the resulting mesh is definitely less noisy.  Sounds like some have found otherwise?

I suspect that while the Nikons images are not pin-sharp at the pixel level, perhaps overall they still contain more detail than a sharp 18MP image.

Of course, to make it interesting, the resulting textures from the Nikons aren't as sharp or color-accurate as the Canon, on the same geometry.  And great textures is a big reason we're all here...

Would it be madness to attempt an array of mixed Canons and Nikons, using all images for the mesh solve, and only the Canon images for textures?  That makes my head hurt. 

[Infinite]

Lee/all, one more question about the Nikons for y'all if you don't mind!  I dig the Canons.  However I've done several tests with 6 each of Canons and Nikons stacked right on each other in stereo pairs, and the Nikons generally result in 20-30% more points resolved, and the resulting mesh is definitely less noisy.  Sounds like some have found otherwise?

I suspect that while the Nikons images are not pin-sharp at the pixel level, perhaps overall they still contain more detail than a sharp 18MP image.

Of course, to make it interesting, the resulting textures from the Nikons aren't as sharp or color-accurate as the Canon, on the same geometry.  And great textures is a big reason we're all here...

Would it be madness to attempt an array of mixed Canons and Nikons, using all images for the mesh solve, and only the Canon images for textures?  That makes my head hurt. 

I've found the opposite. The Canons are definitely superior in every aspect, apart from build quality of the casing and design.

It all depends how you light your subjects. If you use strong, harsh front lighting, Nikons are good. If you use globally diffused bounced light, you will see the Nikons really don't cope very well at all. We see soft, noisy images. They all (the ones I tested) seem to suffer from a yellow / orange colour tinge.

Canon's are just a great average, just right for photogrammetry.

The best thing to do is go with what you feel is best. There is NO rule book with Photogrammetry, so we can all make our own paths of discovery