4:4:4 12-bit Uncompressed DVX100

[Rodger Marjama]

Well its nice to see some physical results none the less. The noise seems sporadic but groups more densely at bright/dark junctures.

I opened it in photoshop to view at 400 percent zoom as I was just doing the same thing yesterday with the same chart shot with the same camera -- but without the aid of your device. Very remarkable difference without any doubt. I can see no compression artifacts in your frame grab at all and the noise I hope is just that "noise" not damaged pixels. This could be verified very easily though, just grab 2 separate frames and compare noise locations. If they're at different locations it's noise if the same it's bad pixels in the CCDs.

Keep it up and prove all us doubting thomas' wrong -- I'm serious.

-Rodger

[Obin Olson]

wow...very cool man very cool!!! :) why is it black and white? looks like it is very clear indeed!!

[Obin Olson]

black and white because it is before the color filter?

[Juan P. Pertierra]

Since it's only from one CCD, it is interpreted as a monochrome image...you can interpret it as red only but there's no point without the info from the other two CCD's.

[Jason Rodriguez]

Hey, These images blow up very nicely to 720p :-)

Also it does appear that you're getting more dynamic range than the camera typically puts out after gamma correction, etc. Additionally, it should be even better when you get that extra bit that accidently got left off. I'm not sure if that extra bit accounts for the slight clipping in the highlights or not, but nevertheless, it's a really nice image.

BTW, you still might want to consider 10-bit DPX log files for encoding these things to get the full range of the 12-bit image into a 10-bit image file. After Effects can read them, so there shouldn't be any compatability issues.

[Juan P. Pertierra]

Well, i think there has to be a single intermittent connection, because there seems to be a 'method to the madness' as far as the noise goes...note how on the grayscale bar on the left, the dots are concentrated in one scale...it seems one of the higher order bits is intermittent, and when it is used for a tone it defaults to a dark color since it is 0....

This is gonna drive me nuts now.

Juan

[Milosz Krzyzaniak]

Juan. I must tell this remains me of a scene in this film - "Hunting for Red October" (or something) and this guy on the submarine trying to establish back communication with their headquaters. As we see we have here the same suspense:)

And now some suggestions.

I would suggest to prepare a special testboard to actually emphasise the issues that we are talking about. So, on one hand there will be a difference between 4:2:0 (4:1:1) and 4:4:4. I think the best way is to prepare a tesboard with some very thin COLOR lines (with sections graduating in their thickness). On this board you have some tonal graduation, and there are some lines with thickness graduation as well. I cannot see it in thsi monochrome still, but I suppose they are black.

If so, they are useless as the luminance is not degraded in any way - we will not see any difference between 411 and 444. So they should be red, green or whatever.

Next thing, this time I would like to ask.

Is 12-bit thing means:

1) the "tonal range" of CCD 12-bit and DV 8-bit output is THE SAME, just there is more "tonal steps" in this range - so the white point and black point are the same

2) the tonal range of 12-bit is somewhat wider than the DV 8-bit output, so we actually can obtain from 12-bit image some tonal information that is not visible in dv (for example some details in highlight and shadow areas beyond the white and black points of DV footage)


If the 2th ver is correct, I would have additional suggestion.


PS. Is TIFF 4:4:4 medium?

[Mark Grgurev]

Forget 720 hi-def!!! That image looks good up-rezzed 1080!!!

[Stephen van Vuuren]

When Juan succeeds with this, the image should look better than HDV due to progressive quality of the CCD's, improved color and lack of compression.

It will interesting to test it head to head with HDV cams since I agree, it should uprez to 720 and 1080 quite nicely.

[Juan P. Pertierra]

Actually, let me clarify something...i guess we're talking about a 4:4:4 signal, because we are capturing RGB but i think the x:x:x notation is associated with YUV signals. an RGB signal per-se is '4:4:4' because it contains the full color information...the only difference between an RGB signal and a 4:4:4 YUV signal in terms of quality is that the YUV has less reduntant information...i'm stretching it here if this doesn't sound right someone please correct me.

Thus, i think TIFF, like BMP is an interleaved RGB (uncompressed in this case) format, so in essence it has full color information, aka 4:4:4. I think the only formats that handle YUV decimation are compressed video formats like DV, might be wrong.

This morning i found an easier way to do the entire wire hookup procedure, i went to the local electronics store and gots some headers and ribbon cable...this should make everything a snap. I'm going to (getting paid for)work for a few hours and then come back to finish the (fun to do)work.

Juan

[Jason Rodriguez]

<<<-- Originally posted by Milosz Krzyzaniak :
Is 12-bit thing means:

1) the "tonal range" of CCD 12-bit and DV 8-bit output is THE SAME, just there is more "tonal steps" in this range - so the white point and black point are the same-->>>

Yes, the white and black points are the same because white is when the CCD's get saturated. What greater bit depth allows is the ability to "move around" within that white and black area and still get a decent picture with enough bit depth. For example, in 8-bit, you've clipped to white, so you decide to underexpose a stop to get the highlights down. Now in post you'll have to to re-normalize the picture so that it looks properly exposed, but keeping the whites in, so you'll use a levels or some other curves filter to modify the gamma some-what. The only problem is that since you've underexposed, you no longer have enough information per pixel to get smooth gradations-so you get banding, etc., because instead of using the entire 8-bits to describe the image, your undexposure limited the scene description to, lets say, 6-bits. Not good. A 12-bit image on the other hand, even if you underexposed and didn't use the two top bits, still gives you 10-bits to work with. So you can then normalize the picture without any problems. Also because of the greater amount of tonal gradations possible in 12-bits, the point at which a scene will clip to white will be further up the "white" scale, due to the encoders ability to describe the information in the whites. In a scene that has a large area of clipped "white", there is actual detail in the real-world scene. With 12-bits, instead of the whole area going to white, only a small portion, or maybe none at all will actually clip to white-it'll be bright, but there'll be some information there because the system had the headroom to describe what was in those whites. So yes, the white and black points remain the same in an absolute sense, but the ability of the system to "see" what's in those very bright and dark areas will be greatly increased because it's not clipping due to a lack of bits to describe what is in the extreme portions of the image.

The important thing about 12-bit capture is that you have the file format appropriate for capture available to you. Trying to cram a 12-bit file into a 16-bit file format is overkill, because there's 4-bits that you aren't using, and when bandwidth is at a premium, that can be a problem. Also taking a 12-bit linear file and clipping off the top two bits to create a 10-bit linear file, or compressing it down to an 8-bit file also does damage. That is why Kodak and SMPTE created the 10-bit LOG file format, Cineon or DPX, that using a logarithmic curve description. It's able to pack the bits from a 12-bit file into the most efficient means possible, while maintaining the bandwidth-friendly aspect of 10-bit files. The logarithmic nature of the encoding places more bits at the bottom of the scale in the blacks, and less in the whites. This is due to the way the eyes see light logarithmically-if you light a match in a dark room you'll automatically see it-it will appear bright, but if you light that same match in a sun-lit room, you'll hardly see it. So for encoding these 12-bit RAW files, probably the most economical and standards-based file format that we could use would be 10-bit Log DPX files, and you would simply linearize them in After Effects using the cineon conversion utility, or with a LUT in combustion.

[Ben Syverson]

With all due respect, I totally disagree. Creating a 10-bit DPX/Cineon log file would require reprocessing the image to fit into the log color space (and probably correct for "white balance")-- which probably can't be done in real-time while capturing.

The ideal situation would be to capture to raw 12 bit files, and then develop a converter application which could export 16bit TIFFs, DPX/Cineon, OpenEXR or whatever you like...

- ben

[Eduardo Soto]

Hi everyone, I admit to lurking and following this thread for a long time. First off, I just want to say how great it is that someone like Juan is out there on the boundaries, pushing the envelope. Thank you!!

Now my first question is, and yes I am purely a creative filmmaker not technical in the least so bear with me, what kind of set up would be necessary to use this mod on location?

I shot a shot a feature on the CineAlta last year and i'm intrigued with the idea of using this mod to shoot pickups ( as opposed to spending thousands more to rend the Cine, camera pretty much ate the bughet -though it was worth it).

How could this mod be practically applied in the field?


thanks so much,


es

[Juan P. Pertierra]

>>How could this mod be practically applied in the field?

Eduardo:

I have a 'production' version of this mod designed, which would be implemented as a box that mounts on the bottom of the camera, and has a firewire 800 output. You can then hookup the firewire to a portable drive, or a computer to capture the raw frames. The camera is of course closed and operates completely as normal.

The mod does not require any soldering or permanent modification of the camera, but it does need to be opened to be installed.

Juan

[Eduardo Soto]

Great! I was worried I' d have to lug a desktop with me!

What kind of portable drives could I use? Or what kind of a laptop setup would be fast enough?

Thanks for responding so promptly,



es