PDA

View Full Version : Re: True 4:4:4? Log Curve? Crop Factor?



ectobuilder
06-13-2012, 05:41 AM
Is this camera a TRUE 4:4:4 camera? Or is it taking say a 4:2:2 colour space and then outputting it as a non-destructive RAW file for video?

And does it have some kind of log curve built-in to increase dynamic range? Is that where the spec'd "13 stops" is coming from?

And what is the EXACT crop factor of this camera?

And what happens if the built-in battery loses it's ability to be recharged over time? Does the warranty cover it? If so, for how long?

laco
06-13-2012, 06:01 AM
Firstly, I saw you open up some new threads, but it could be wise to search in the forums about these:)
I'll try to answer these:

Is this camera a TRUE 4:4:4 camera? Or is it taking say a 4:2:2 colour space and then outputting it as a non-destructive RAW file for video?
4:4:4 is Chroma subsampling, not color space.
This camera - currently - supports three recording formats. Two of which is 4:2:2, and there is RAW which means it doesn't 4:2:2 or 4:4:4 (it is before chroma subsampling), so it has all the data for being 4:4:4. So yes, this camera can output 4:4:4 at the end. You can search for Barry Green's posts, which details that even if a camera records RAW (that can be output to 4:4:4), Bayer sensors has a limitation to truly recording color data to every pixel. But this is what most camera (RED, Alexa) has, so don't fear:)


And does it have some kind of log curve built-in to increase dynamic range? Is that where the spec'd "13 stops" is coming from
The sensor has a 13 stops DR according to blackmagicdesign. LOG just ensures that you won't loose detail because of a gamma curve.
Check the log pictures on John Brawley's blog:
LOG: http://johnbrawley.files.wordpress.com/2012/04/leah_1-20-2.jpg
GRADED: http://johnbrawley.files.wordpress.com/2012/04/leah_1-20-1.jpg



And what is the EXACT crop factor of this camera?
If you compare it to Super35mm size, then 1.6x
If you compare it to photographic full frame, then 2.3x


And what happens if the built-in battery loses it's ability to be recharged over time? Does the warranty cover it? If so, for how long?
If it fails during warranty, I'm sure it will be replaced.
If you need to completely change it BMD will do it for you for $80. Quite cheap if you ask me.
Also - I think - nobody will shoot with the internal battery. This serves as a tandem power source not to lose power when you're changing external batteries.

ectobuilder
06-13-2012, 06:44 AM
So you're saying that RAW is above 4:4:4 in terms of preserving color data coming off of the sensor? In other words, even 4:4:4 is losing color data?

laco
06-13-2012, 07:10 AM
So you're saying that RAW is above 4:4:4 in terms of preserving color data coming off of the sensor?
RAW isn't a Y'CbCR signal where you have a sense of Chroma subsampling. It's before the color subsampling process.
But if you convert RAW to Y'CbCr signal, you can get 4:4.4 out of it.


In other words, even 4:4:4 is losing color data?
No, 4:4:4 means that no color data is thrown away.
But the de-bayered color data that's coming off the sensor doesn't have full chroma resolution.


If you want a simple-simple - but flawed in the details - answer, yes you can have 4:4:4 color sampling coming out from the BMC by recording in RAW.

nickjbedford
06-13-2012, 07:18 AM
But the actual resolution of the colour itself as a side effect of the Bayer sensor pattern won't be pixel for pixel 444.

laco
06-13-2012, 07:49 AM
But the actual resolution of the colour itself as a side effect of the Bayer sensor pattern won't be pixel for pixel 444.

I don't like to mix-up the two:)

1. part: Yes, camera records RAW, so after chroma subsampling, you can get 4:4:4 out of it.
2. part: This camera has a Bayer sensor which results that if you don't downscale your source resolution, you will miss some chroma data. But this can be confusing for a lot of people, and they would think that this camera has some disadvantage, but the truth is that is is true for all Bayer cameras.

Make it simpler, if you ask what a RED EPIC is capable of, they will say it can record 5K RAW, which you can "convert" to a 5K 4:4:4. Even though there won't be full chroma information. Because this happens to all Bayer sensor cameras, we can leave out this from the chroma subsampling subject IMO.

Also.. sometimes 4:2:2 is still very good. You eyes won't really notice the difference between 4:4:4 and 4:2:2 (unless you do some 1600% zoom-in), you will only have some difference when you start to key some things in Davinci.

Tom
06-13-2012, 08:38 AM
Personally I favour this analogy - Film stock as no actual resolution, you can subjectively compare it to a digitised image which has a finite resolution - but the actual resolving power of film is not a finite amount. RAW data with a bayer pattern has no real resolution as such, some raw converters allow you to oversample at the first step for example, until the RAW data is debayered and quantised into a non raw format - it is similar to analogue data (not literally of course - this is just an analogy) Depending on how sophisticated the debayering process, will depend on the quality of the processed image. The image can be processed into a 4:4:4 format or 4:2:2 or whatever. In reality, the difference between the actual quality of the 4:4:4 image compared to other 4:4:4 images of the same resolution comes from the resolution of original RAW data and the resolution of the converted 4:4:4 footage.


So if we take this raw image without it being debayered: http://blogs.elphel.com/wp-content/uploads/2010/10/source-bayer.png
The quality of final image is very much dependant on how well it is debayered, look at these two examples of two different debayering processes and the aparent resolving power of each image:
Mode1: http://blogs.elphel.com/wp-content/uploads/2010/11/deconvolved_with_threshold.png Mode2: http://blogs.elphel.com/wp-content/uploads/2010/10/gaussian1.png

The apparent clarity of those images is quite different, but they both come from the same raw data.

We could use a really crappy debayer algorithm for the BMC and have very soft blurry but fully 4:4:4 footage, or we could use a really good debayer algorithm and have nice detailed sharp but fully 4:4:4 footage. OR we could use a sensor with an even higher resolution, and downsample it so that when its converted to 4:4:4, there is more information per pixel and as a result, the 4:4:4 image will have a greater resolving power.

A lot of what i have just said is not TECHNICALLY accurate, but it is just me trying to help to conceptualise the difference between the two. Debayering and Chroma Sub-sampling are completely different things, but they both can have the same impact in the end result in terms of resolving power. If we pretend that FULL HD is our final format, that the more the sensor has a higher than HD resolution, the more the HD footage will appear to have greater detail. 4:4:4 is just a way of saying that no chroma sub-sampling has occurred. It does not mean that the camera has a sensor which oversamples to resolve 4 pixels per final pixel.

I apologise if i have made any massive mistakes in this post, I am just trying to help explain and clear any between this 4:4:4/RAW/Debayer debate which seems to frequently crop up.

tldr version: 4:4:4 simply means no chroma Sub-sampling has occurred - which is true of the BMC raw video format.

laco
06-13-2012, 08:50 AM
tldr version: 4:4:4 simply means no chroma Sub-sampling has occurred - which is true of the BMC raw video format.

yep:)

vealti
06-13-2012, 03:04 PM
I thought this article by Jay Holben gave an excellent yet simple explanation on what raw and 4:4:4 is, and how the Bayer sensor pattern effects things.

http://www.creativeplanetnetwork.com/tags/raw-imagery/dv101-raw-deal-what-does-it-mean-record-raw-imagery/59526

VampireZIM
06-17-2012, 10:11 PM
So, I am going to be doing a lot of keying. I am assuming that this camera will do much better with keying than any DSLR, is that a correct assumption?

nickjbedford
06-17-2012, 10:38 PM
So, I am going to be doing a lot of keying. I am assuming that this camera will do much better with keying than any DSLR, is that a correct assumption?

This camera will generally produce a much better image period, disregarding the image content itself (such as depth of field and such).

Philip Lipetz
06-17-2012, 11:21 PM
So, I am going to be doing a lot of keying. I am assuming that this camera will do much better with keying than any DSLR, is that a correct assumption?

yes

laco
06-18-2012, 01:11 AM
So, I am going to be doing a lot of keying. I am assuming that this camera will do much better with keying than any DSLR, is that a correct assumption?

Yes, but if you've worked with DSLR footage this will be a bit different.
DSLR compression is smoothing out details, so when you try to key something you will get a smooth enough key, but it wont be precise.
If you key a better footage, you will see a more precise key, but it won't be as "smooth". You'll have to "blur" the key itself, but after that you will have a much nicer key:)

Also for keying the uncompressed recording will truly shine.