Implementing Camera Model and Sensor sizes

[Jon Bragado]

Hi,

I´m Jon and i´m a camera td with expertise in the realm of matchmoving. I have long been waiting for a database like this, that would stop us from chasing people to get camera information from, specially in commercials world. So, thank you for that.

As important as camera information and lens information is, the key to everything is the filmback in film cameras and sensor sizes in digital cameras. In my honest opinion this could look trivial but it´s not. 

Depending on the capture format the area of the sensor used is different, an example of the variations for one single camera is this.

http://www.arri.de/fileadmin/media/arri.com/camera/Digital_Cameras/ALEXA_XT/ALEXA_XT_Recording_Areas_Surround_Views_and_Framelines_SUP_8.0.pdf

Wouldn´t it be great, as we fill the database to have an option to choose the corresponding sensor size depending on the shooting format? Not all the plates are shot at the same resolution and sometimes not even with the same camera; specially when more than one unit come into play.

It could be a nive feature to have.

Also could be great to have a box where we could choose the overscan value in case footage needs to be undistorted and perform a calculation that could give us the final undistorted resolution for that plate. This is done by multiplying the overscan by the original resolution of the plates. That way, exists something which a camera td can compare to and be sure the undistorted resolution is correct.

Well, i throw the suggestion and expect some comments.

Many thanks,

Jon,

[Sam Richards]

On Fri, Dec 13, 2013 at 5:01 AM, Jon Bragado <jonbr...@gmail.com> wrote:

I´m Jon and i´m a camera td with expertise in the realm of matchmoving. I have long been waiting for a database like this, that would stop us from chasing people to get camera information from, specially in commercials world. So, thank you for that.

How common is it that you cannot derive the film-back size (and I'm assuming you are trying to determine the sensor size in mm?), from the capture resolution? I have seen issues like this for animorphic vs non-animorphic values. Or is somebody cropping the element before its delivered to you?

I'm also not quite understanding the overscan distortion, surely that is just a cropping value?

You are starting to get into areas where I'm getting pretty rusty, so I apologize if I'm missing something.

Sam.

[Charles Clavadetscher]

I doubt people on set (Camera crew, DP etc) will have millimeter accurate specs.

Even if they had the numbers, I would want some 3rd party, reliable verification.

This sounds like a whole new project: a camera company official spec list (per camera), with the possibility for users to add custom cameras or variation when they use cameras not on the list, or do the work to derive the film back themselves.  

It would need to cover standard "A" list movie cameras from Arri, Panavision, Red, etc, and hopefully include less common cameras such as various still cameras and some consumer and prosumer cameras that are often used as witness cameras.

These need to be match moved too.

Of course we can't do every camera on the market, but we can get most Canon and Nikon specs, and probably get a good list of cams from users. Maybe the Sony F5 and others?

Perhaps the VES would like to make a document on how to calculate your camera's film back? [yet another project].

charlie

On Fri, Dec 13, 2013 at 5:01 AM, Jon Bragado <jonbr...@gmail.com> wrote:

--
You received this message because you are subscribed to the Google Groups "ves-tech-camera-reports" group.
To unsubscribe from this group and stop receiving emails from it, send an email to ves-tech-camera-r...@googlegroups.com.
For more options, visit https://groups.google.com/groups/opt_out.

[Sebastian Sylwan]

I agree with Charles, however, as the spirit of this project was to provide a common and simple way to input data that is easily available to the crew to encourage its input.

I doubt the sensor size / exosed area data would be available to the crew, and given the variability, it should not be calculated on set. I'd rather have reliable camera and lens information, on which to base that calculation on.

However, recording format and camera resolution (as set on the camera) may have a place here.

S

__
Please excuse typos, fingers too big for BlackBerry keys.

---

From: Charles Clavadetscher <charl...@gmail.com>

Sender: ves-tech-ca...@googlegroups.com

Date: Sat, 14 Dec 2013 12:11:25 -0800

To: <ves-tech-ca...@googlegroups.com>

ReplyTo: ves-tech-ca...@googlegroups.com

Subject: Re: Implementing Camera Model and Sensor sizes

[Charles Clavadetscher]

A)  Just in case my comments came across as harsh, it wasn't meant that way.

It seems like like film back data isn't something in a data wrangler's responsibility or control.

B) I just re-read my email and wanted to clarify one item.

When I suggested a list of film backs " ... per camera,"  I did not mean individual cameras 

on set or the same model on a different production, but rather "same make and model".

It would be up to someone, us maybe with help from the camera vendors, to make sure changes to chip or other image tech is up to date, and new models of the camera get assessed and included as needed.

charlie

[Aaron F. Ross]

Of course the production crew doesn't have the correct aperture data handy. The point is that it's critical information, *no one* has it handy, and someone needs to look it up. Even if it's not possible / optimal for production crew to input aperture size, I think it should still be on the spreadsheet. Matchmovers can then enter the relevant data in post. It just makes sense to have all of the information in one place.

[Sam Richards]

Back to Charlies proposal where we potentially have a central place to map the Camera format/resolution to sensor size, and provide it in a way that its easy for people to either download it or query against it (e.g. provide a google-spreadsheet, as well as a json url).

I'm assuming for the digital cameras that there shouldn't be too much variation?

I remember Charlie needing to do it for particular scanners, but I'm only proposing doing the digital cameras. Are there other things that can cause variations? e.g. particular lenses? 

Also is this something where manufacturers are willing to give out this type of information freely (Arri does seem pretty open, but red is much harder to find this type of thing)? 

Also, does it require actual measurement? Or can we go off specs?

As for the one place for everything, while I like the idea, I don't think that's possible, we really need to assume that you are already merging multiple data sources (e.g. meta-data from the camera with the camera-reports), so assuming you can combine it with additional sources doesn't seem a stretch. However, I'm open to coming up with ways to make this easier to do, possibly developing python libraries that can help with that.

Sam.

On Wed, Dec 18, 2013 at 1:22 PM, Aaron F. Ross <aa...@digitalartsguild.com> wrote:

Of course the production crew doesn't have the correct aperture data handy. The point is that it's critical information, *no one* has it handy, and someone needs to look it up. Even if it's not possible / optimal for production crew to input aperture size, I think it should still be on the spreadsheet. Matchmovers can then enter the relevant data in post. It just makes sense to have all of the information in one place.

--

[Kirk Chantraine]

Figured I'd chime in with my two cents.

I think it would be beneficial to track but I think the question of which data lives where does come up. If you have the camera model I'm thinking much of it could be looked up in post as mentioned.

Also, does it require actual measurement? Or can we go off specs?

Measurements will be key, but they may be included in the specs.

Some searching online can provide some good resources though, for anything not found in the camera metadata:
http://www.reduser.net/forum/showthread.php?66210-Scarlet-X-Data-Sheet
http://www.dpreview.com/glossary/camera-system/sensor-sizes

We're having to do a lot of digging on this right now as we're dealing with some plates shot and only in-camera data. Will echo thanks on the template database though - definitely pushing for its use!

[Sam Richards]

Thanks for all the links, I started putting a google-spreadsheet together, see:

https://docs.google.com/spreadsheet/ccc?key=0AnE1uazTlVCGdHRXdmdKcG1oSm1SS3dvSG5PRlptMVE&usp=sharing

Can you take a look and let me know if the columns make sense? I do see quite a few places where the diagonal is also mentioned, I've actually left it off, since I'm assuming its obvious. 

I'm happy to keep adding to this, and publishing it in a variety of formats (csv, json, etc) as well as allowing direct access, if it makes sense.

In terms of verifying the values, how would we do that?

sam.

[Sam Richards]

Hi, I still have no feedback on this. My preference is that if we add a field to the camera-report format for recording-resolution/format, I'm hoping this could be a slate field, and not per take.

I can see there being some odd cases where this might vary per take, but I would hope we can assume this is an extreme case that could be covered by the notes field.

Please let me know...

Sam.

[Sebastian Sylwan]

Makes sense to me. 

I think one of the key factors for acceptance will be adhering to KISS as much as possible. 

S

[Kirk Chantraine]

Yep, looks good! Agreed regarding per-take coverage. Thanks Sam.

[Jon Bragado]

Well, It is quite common that they shot say with a RED at 5k, with a 50mm lens. Then you find the plate is ingested at 2k. How much of the sensor is used yes, you can determine it by the theodolites theory, stablish the croping factor and then recovert the lens, obviusly a 50mm lens shot at full sensor 5k, size isn't the same at 2k. 

Then it comes the anamorphic, which is a problem, with the squeeze factor, which by the way is not always 2.

All I was saying is to have a pulldown menu where you have a camera models list and a place to pick the resolution it was shoot at. Because the ingesting normally is a downrez version of the original.

If a matchomver sees that in the report sheet, he knows that it was shot at 5k, so it's the highest resolution of the camera which means it was using the full sensor. Then he can do his math as we all know.