Possible solution to color correcting / exposure for 5D Post Production

[Arsen Avi Romanovich Bortnik]

Correcting footage for the 5D 4:2:0 post work is very limited, but I may have a work around; this is theoretical at the moment, but wanted to see if anybody had any thoughts. I am not sure how FCP or other non-linear editing programs execute color correction, but I think color correcting a single frame in Photoshop out of FCP or AE may have higher correcting capabilities without damaging the frame. In the past for rotoscoping purposes, and other effects I did with my projects I would export 20 seconds of footage at 24 fps x 20 would = 480 frame/pictures. The frames export out of After Effects in sequential order as well FCP. I would do my effects frame by frame, than import them back into FCP, and you have the same clip with a Photoshop work-around. For this situation you would apply an effect to one frame of the 5D footage; if it would be color correction; after correcting one frame, you would apply the same effect to the rest of the frames via a batch execution. I estimate that each frame out of a 5D is roughly 4MB, compared to Red Raw Code at 5K which is roughly 30MB. Of course it is simple to go to your 5D footage in FCP 3 way color correct and be done with it, but the matter of fact is we are in an industry where we take pride in our work, and any work around with our art is truly worth it. Back in the old days in order to colorize an original B&W film they would do it frame by frame, roughly 90k frames, sometimes much more. In conclusion, just wanted to see if anybody had any thoughts on color correcting in i.e. FCP vs the proposed process of Photoshop, is there possibly an advantage in doing it in Photoshop, with a greater forgiveness, and tolerance without changing the integrity of the 5D footage. I am not talking about important sequential frames of your whole project; say 40k frames for a short film, but just the scenes where you might need it most.

[Phil Rhodes]

but I think color correcting a single frame in Photoshop out of FCP or AE may have higher correcting capabilities without damaging the fram

e

What makes you think that?

There's no reason to believe that Photoshop uses any better mathematics than Premiere or Final Cut; in fact, Premiere now uses the same 32-bit floating point code as After Effects, and does a very nice job. I would assume Final Cut was similar unless someone told me otherwise. What may help you is doing a better interpolation from the compressed camera original file; google for "5DtoRGB" for an idea on how this can be done.

I estimate that each frame out of a 5D is roughly 4MB, compared to Red Raw Code at 5K which is roughly 30MB

Where d'you get those numbers? Are you talking about compressed or decompressed sizes? There's obviously no way a Red can record at anything like 30MB per compressed frame onto a compactflash card, the best of which barely do twice that a second.

Uncompressed, in 10-bit, a 1080p video frame is about 7.5MB; a 5K frame would be about 50MB, but in both cases the lion's share of the data is mathematically generated interpolation in the case of Canon or Red cameras.

P

[Arsen Avi Romanovich Bortnik]

I can't find the post that I have read in either red user or on the actual red site itself, stating that each redcode raw frame is 37MB roughly, this is just what I read, if I stumble upon the article I will reply a post, of course you gotta be careful where you get your info from on the net, but as I recall this was a professional article, not from a anonymous user/blogger. The way I got roughly 4MB for each frame size out of the 5D --> 4GB CF = 12 min of video at 24fps | 24fps x 60 sec = 1440 frames x 12 = 17,280 total frames from a 4GB card | 17,280 / 4GB (4096) = 4.22 MB... actually just thinking about this calculation that I did a while ago, and knowing that the 5D or 7D records 12 min for every 4GB of space, I have to assume they are referring to 30fps which in that case is 5.27 mb a frame. although video compression borrows data info from every last frame, and so on... when it comes to exporting just a single frame out of a given editing software such as fcp, I have to assume that all the color information frame to frame that was compressed via the 264 codec will be in every single frame when you choose to export, where as if you have recode raw or any other raw footage obviously each frame contains all the information from the get-go. I just exported a single frame from 5D footage at it's highest quality jpeg at roughly 1.2 MB and a Tiff at roughly 6MB

 

There's obviously no way a Red can record at anything like 30MB per compressed frame onto a compactflash card

 

I know Red has an option to record onto CF cards, though it is a hectic system, constantly reloading cards and footage, I am not completely sure what you meant by that quote, I have read that they have just released a 128GB CF card that is capable of recording 4:4:4 uncompressed footage, though at $1,200 I hope thats the least it can do, personally when I switch to the Red system I would go for the red mags, or Im sure they have already developed or will be developing other external recorders with greater capacities.

http://www.bhphotovideo.com/c/search?Ntt=SanDisk+128GB+Extreme+Pro+Compact+Flash+Card&N=0&InitialSearch=yes

 

but I think color correcting a single frame in Photoshop out of FCP or AE may have higher correcting capabilities without damaging the frame

 

What makes you think that?

There's no reason to believe that Photoshop uses any better mathematics than Premiere or Final Cut

This was just a theory, and I wanted to get some input on it, thats all, It looks to me that you are more experienced and knowledgable with this topic, so I appreciate the insight!

Edited March 14, 2011 by Arsen Avi Romanovich Bortnik

[Phil Rhodes]

just released a 128GB CF card that is capable of recording 4:4:4 uncompressed footage

 

There are now 2.5" form factor SSDs that will do this quite routinely. I haven't heard of CF cards faster than 100MB/sec; the theoretical limit is 133.

[Nathan Blair]

There's certainly no harm in giving new things a try, but I think Phil is right. If the data isn't there, photoshop unfortunately isn't going to save us. The only way to get a Raw file off of a Canon is by taking a still, isn't it?

 

Arsen, you should check out 5DtoRGB though, it's worth reading about at the least. I just wish they would improve it so that you can encode more than one clip at once!

Edited March 14, 2011 by Nathan Blair

[Arsen Avi Romanovich Bortnik]

For now that is correct, I found a very interesting article in The Cinematographer about the making of 127 hours, they actually used two cameras, I forget which ones, but they basically stripped for example a viper, where they had at one end, the lens and as little of the camera operational ssystem as they could all wired to a control type device with the rest of the camera functions, they used two such cameras as well as what appeared to be two 5.6 HD monitors. There had to be a better solution for the enclosed space that they were working in, either they had no other choice, or placing an robotic articulating arm in that space was costly, or they simply wanted to do hand held work, which in my opinion can't be mimicked from a controller, especially reacting to an Actors real time subtle actions. The most interesting thing that Boyle's Cinematographers tried to do was to strip the 1D and they tried to get the raw format video out of the Canon 1D, and were unsuccessful, if you have seen the film there are some instances where we see Franco in somewhat of jittered, or skip type footage where you only see about every other frame. One of the Cinematographers technique was to use the 1D's fast burst of 10 pictures a second in raw, and basically Im assuming dropping the sequential shots into the editor, giving it a 24 fps along with the film, its def not flowing 24fps footage but its a pretty cool effect which can be corrected in the Raw format, of course with D Boyle's films, he does a lot of experimental shots, split screens etc, Iv'e always found such sequences undesirable, but have recently learned to love all aspects of Directors film making; I have to note that I am a recovering OCD Perfectionist, lol... You still have to do excellent work in everything you take on, but when you are a perfectionist, what may seem not perfect to you just may be of opinion, and actually look great.

 

I really wanted to try to accomplish some sequences in burst mode, but with the 7D's 8fps mode, and then tripling it to 24, giving it that skipping effect. Only thing; I don't want to mess with the integrity of the camera's actuations, I rarely use my 5D for photography, and have read that dslr's life cycles, go any where from 75k to 200k and the camera's cooked.

[Phil Rhodes]

Hacking DSLRs for their raw video is a tricky one, because the CMOS sensors will include a lot of onboard analogue and digital electronics. With CCD cameras, it's usually quite obvious where the RGB analog-to-digital stages are and they're often off the shelf ICs that can be persuaded to do other things; this is how the Andromeda mod for the DVX100 worked.

 

The thing is, even if you could get raw video out of a Canon DSLR, it'd still have appalling aliasing issues, and probably even more obviously given the sharpness of an uncompressed recording.

 

If you are desperate to do this, you get a Panasonic GH2 and record its HDMI output, then decimate all the duplicated fields out later on.

 

 

P

[Arsen Avi Romanovich Bortnik]

Copy that, it's not something I'm interested in, I know there have been slot of talks about it, plus I can only assume getting a larger raw format or even a newly based 2k to 3k or whatever the filetype may be out of the 5d will probably heat up the sensor even quicker, that would be great though, to have a fully uncompressed resolution out of the 5d, a euphoric state of mind for us all. I appretiate the input on the GH2 my friend! Will actually check that out now. A full raw resolution at the highest 5k by 3G roughly would def be great to eliminate the Moire, that that's not even in the question, I work with a smallhd as well as a vizio 22 inch razor LCD to make sure I have no moire patters, if theyappear I simply go to shallow dof, for shots that don't call for a shallow dof artistically I try to get it shallow enough to not have a crazy dof where I need to elistrate emphasis on a scene, for example rack focus from an important prop to character