H.264 vs RAW - Canon 600D / T3i

gebo

gebo

Made a quick test comparing both formats. This was mainly done to complare the dynamic range.
The flicker on the h.264 clip is caused by post processing (my mistake).

What do you think?

https://www.youtube.com/watch?v=Eng6lHjFS-g
 
Sort by date Sort by votes
Looks good!

I'm so excited by this new development! If the T3i has this function now, I'm so hoping it comes to the T2i. Definitely considering getting a Mark III now though.
 
Upvote 0 Downvote
Upvote 0 Downvote
iv seen the Raw comparison to h.264 on t3i with many shots exampled, the fact they have to stretch the raw ratio to fit the h.264 ratio and it still comes out sharper and cleaner is insane!

i think il wait a few more months before testing, im not in a position to replace my camera if i fry it.. which with raw on this camera is a risk im not willing to take until further lab rats have been disposed of :)
 
Upvote 0 Downvote
Zensteve said:
The few recent posts I've seen have all been talking about the T3i as the base. Is the T2i out of luck with this new RAW development?

.
Click to expand...

I would imagine T2i and T3i have similar limited speed capabilities. Seems to be talk around the forum that it might be possible to have full-res compressed raw out of such cameras, but whether or not that will come to fruition who knows.

I'd be happy to see ProRes 422 files out of DSLRs, let alone raw. From what I can gather, T3i raw is pretty much limited to 640x480 cropped sensor mode, so whilst you're getting a slightly larger than 720p raw resolution, it seems to be coming from the sensor crop mode, meaning you're gonna have to use super-wide lenses...

Also, the difference between 1080p H264 and stretched raw seem to be most prevalent in DR, not as much in sharpness or clarity. Also seems that the difference in DR between raw and Cinestyle is apparent but not huge - definitely would be worth giving this hack a shot when it's eventually finalised, but not sure if it's 100% worth it over cinestyle, at least for now on the T3i.
 
Upvote 0 Downvote
I don't see how ProRes would be huge since you can convert your H264 to ProRes (although I don't see what you would, ProRes is overrated).

RAW on the other hand, you can't convert to that.

The question is why is there more Dynamic Range in RAW ? And why can't you shoot 1080 with RAW ?
 
Upvote 0 Downvote
Zensteve said:
The few recent posts I've seen have all been talking about the T3i as the base. Is the T2i out of luck with this new RAW development?

.
Click to expand...

They are focusing on the T3i right now but say that it may be possible on the T2i too.

TheArtist said:
The question is why is there more Dynamic Range in RAW ? And why can't you shoot 1080 with RAW ?
Click to expand...

There's more dynamic range in raw... becaus it's RAW, no compression, so more color information per pixel.
You can shoot 1080 RAW on the 5D mkIII because it has a high write speed (RAW files are huge...) ont the 600D / T3i you're limited to a write speed of 21 Mb/s which isn't enough for 1080 RAW so you'll have to stick to lower resolutions...
 
Upvote 0 Downvote
TheArtist said:
I don't see how ProRes would be huge since you can convert your H264 to ProRes (although I don't see what you would, ProRes is overrated).
Click to expand...

Because ProRes is much less compressed than H264, so as an acquisition format, you're gaining more resolution, and potentially more DR, depending on how it's implemented.
I don't really know the deep mechanics of these cameras, so it's hard for me to say how it could be implemented, but doing so would definitely give you an edge over H264.

If you want, you can analogise it to photos:
When you take a photo in raw, you're getting all the information the sensor sees, and all the processing is left for you to do later. That's more or less the same thing that happens in camera raw (except you're essentially capturing 24 raw images per second, so you need a fast drive - in the T3i the write speed is not fast enough to sustain full 1080p).
Shooting in ProRes is I guess kinda like shooting in .png - you still retain a whole lot of information, but not all information. ProRes 4444 even includes an alpha channel for transparency, similar to the way a .png keeps transparency.
Shooting in H264 is kinda like taking a photo as a .gif, that then re-saves to a .jpg at 1/2 quality. You know how when you save an image to a .gif you lose a whole bunch of the colour and resolution?

Of course, the video formats are not equivalent to the photo formats, but it's perhaps easier to explain in a simple way than having to get into de-bayering and compression ratios and all the rest of it.


Also, converting your H264 footage to ProRes makes editing it easier on your computer. I posted in a different thread about the specifics of why this is so, I'll see if I can find it.
 
Upvote 0 Downvote
from what iv read, they are looking at killing off all the non essential features so that they can improve the frame size, we are in exciting times for Canon DSLR's.

the main question for me now is this, with the new hack and surely better quality to come, does this mean that the gh2 is now in trouble or at least challenged?
 
Upvote 0 Downvote
Has there been any test to prove that ProRes capturing gets you a better image than h264 ?

jax_rox said:
Also, converting your H264 footage to ProRes makes editing it easier on your computer. I posted in a different thread about the specifics of why this is so, I'll see if I can find it.
Click to expand...

Not true on Premiere (or maybe by a small margin).
 
Upvote 0 Downvote
8salacious9 said:
from what iv read, they are looking at killing off all the non essential features so that they can improve the frame size, we are in exciting times for Canon DSLR's.

the main question for me now is this, with the new hack and surely better quality to come, does this mean that the gh2 is now in trouble or at least challenged?
Click to expand...

The GH2 won't be challenged by the 600D/T3i in terms of image detail/resolution.

But the 5D mark III is a different story, now with the hack it's as detailed as the hacked GH2 (but still more expensive...):

GH2 vs 5d mark III RAW comparison:

https://www.youtube.com/watch?v=SFBc7i1Q_0w
 
Upvote 0 Downvote
That video doesn't really prove much - if anything it illustrates to me that the differnce between H264 and raw in the Canon 5DmkIII in certain shots is not huge.

Shoot a comparison between the GH2 hacked and the Canon 5DmkIII raw in a shot/scene where there's a big contrast between the highlights and the shadows and you should start to see some difference in dynamic range.
 
Upvote 0 Downvote
TheArtist said:
I don't see how ProRes would be huge since you can convert your H264 to ProRes (although I don't see what you would, ProRes is overrated).
Click to expand...

If you think ProRes is overrated compared to H.264 as an acquisition codec I'd suggest you don't have enough experience with post production to make a judgement call.

It isn't a question of whether you can convert to ProRes or any other format later. Encoding the 14bit data off the sensor to ProRes preserves far more of the original information than h.264 - which is something you can't get back by transcoding the h.264 to ProRes later. It also introduces far fewer compression artifacts than h.264, which will allow you to maintain a cleaner image through multiple encoding generations from post to delivery. Encoding it in camera saves you both the time of a transcode later, and the generational loss that comes from that additional encode.

It does all of that while producing files that are significantly smaller than the raw data off the sensor, giving you longer record times per card and allowing you to use cheaper, slower cards. It's significantly more practical than working with raw, with far, far less compromise than h.264.

TheArtist said:
Has there been any test to prove that ProRes capturing gets you a better image than h264 ?
Click to expand...

ProRes is a significantly higher quality codec than h.264 (especially as implemented in camera).

10 bit, 4:2:2, intraframe-only, low processing requirements, high bitrate but low enough to be manageable in post without exotic hardware. It also holds up extremely well to multiple generations of encoding in the post process; it's designed from the start for post-production.

H.264 is 8 bit, 4:2:0, interframe, low-bitrate - it's really a delivery codec that's been forced into the role of capture by camera manufacturers. It's also very computationally expensive. That's become less of an issue on the post production side due to GPU acceleration, but it's still a concern on the acquisition side as compromises have to be made between performance/heat/battery life in a camera body. Canon's DSLR implementation is particularly weak due to this, which is part of the reason there's so much to be gained from raw capture.

Just the 10 bit vs. 8 bit difference is huge for post production. That's 1024 levels per channel vs. only 256 - that especially becomes a significant advantage when you're shooting a flat or log-like color profile. In that situation with an 8-bit codec you're essentially sacrificing mid-range fidelity for shadow and highlight detail - which can negatively impact important things like skin tones. With 10 bit you get more levels each in highlights, midrange and shadows than you get in the entire 8-bit image. Now compound the issue by adding the 4:2:0 color sampling of h.264 and you are only making a bad situation worse.

Now you asked about a test - the problem is the test isn't as simple as recording the same image with both H.264 and ProRes and then looking at them. There are tests out there from several cameras (internal vs. an hdmi ProRes recorder), and they basically show only minor visual differences. That's exactly as expected - h.264 achieves it's high compression ratios by discarding information which is not visually noticeable, so it looks great to the eye. To really see the difference you need to shoot the same footage with both then run it through a full edit, color correction and encoding for output. The problem is that the information h.264 throws away - which is not visually noticeable - is often useful later in the post process, and can become visible later in your finished piece.

I don't think we're likely to ever see ProRes available through something like Magic Lantern. Enabling raw recording is a process of elimination - skip the on-camera scaling, debayering and compression steps and just write the data to the card. ProRes or another high-quality codec would require adding new encoding code to the camera, which is a much more difficult process. I think it's especially unlikely with ProRes because there's no open source implementation of the codec; if they do add any additional codecs they're likely to be something that's non-proprietary.

My main interest in the blackmagic cameras though is due to their ability to record straight to ProRes. I see raw as a bonus that you might use in specific situations where it could give you an edge, but direct to ProRes is the everyday shooting format that still outperforms all the other compressed camera formats available now.
 
Last edited:
Upvote 0 Downvote
You must log in or register to reply here.
Back
Top