Canon 5D Mark III -IPB contains more detail and has less artifacts than ALL-I

Regarding in-camera sharpening vs. none as well as camera profiles: tests that we can show and share are the best way so people can form their own opinions on the test results (then test themselves).
~2 years ago ALL-I was clearly inferior to IPB: http://www.dvxuser.com/V6/showthrea...more-detail-and-has-less-artifacts-than-ALL-I

When I initially posted-sharpened frames from pre-production 5D3 footage, many folks didn't like that idea and felt the out-of-camera footage should be sharp as is. I emailed Bloom and asked him to try zero in-camera sharpening and post sharpening. He created a video demo doing this and then it was cool, lol.

When I tested with CS6, ALL-I looked OK (no more macro blocking issues vs IPB). Then the question was disk space. For most cases, except fast motion, IPB works great. ALL-I tends to have more noise so that might have an artistic look some may like vs. the cleaner IPB (not clear if Canon is doing more NR before compression or is a result of Canon's IPB implementation). A better IPB implementation would effectively switch to ALL-I during fast motion providing the best of both worlds... Someday :)

Keep in my ALL-I can take 3x the disk space vs. IPB (over 100Mb/s)...

I used to shoot flat but now try to shoot as close to the final desired look as possible (I prefer saturated looks in many cases, though always with a touch of 'unreality'). Less work in post = good in my book :)

I've been doing a lot of tests with ML RAW, and after comparing with Canon's H.264, especially in low light, ML RAW is of course far superior (since I don't need to denoise with Neat Video very often (ACR denoise can be sufficient). Neat Video is great but very slow (compared to the real-time flow of PPro)).

5D3 RAW has a lot more detail than Canon's H.264, however my FS700's Sony implementation of AVCHD (H.264) at a measly 24Mbit/s produces more resolution and detail than 5D3 RAW (Canon C100 has a similar performing 24Mbit/s codec and perhaps slightly more resolution than the FS700 (but none of the slomo, and no Speed Booster / Full Frame look). Matching footage I need to sharpen the 5D3 RAW and blur the FS700 H.264 :)
 
jcs said:
5D3 RAW has a lot more detail than Canon's H.264, however my FS700's Sony implementation of AVCHD (H.264) at a measly 24Mbit/s produces more resolution and detail than 5D3 RAW
Click to expand...

This is not true. I had the FS700 and shot the exact same scene as with the 5d III raw. 5d III BLEW away the FS700. It blew away. The difference wasn't slight, it was huge. 5d was sharper, no compression artifacts, cleaner. In every single category the 5d raw performed better. The fs700 has huge amounts of sharpening applied internally but it looks horrible (some people like those halos though). The 5d had more real actual detail. The scene had a tree, grass and movement. The grass on the fs700 was completely destroyed by the AVCHD compression, all the tree branches had sharpening halos etc. Not to even mention the color difference.

I also find it odd how you've never seen the IPB mush issue. I just can't trust your eyes man.
 
Last edited:
For indoors, low dynamic range, low motion, the higher resolution image from the FS700 is visible vs. 5D3 RAW. A couple of chart tests:
http://provideocoalition.com/awilt/story/high_speed_and_low_light_with_the_nex-fs700/
https://vimeo.com/67090302

Even in RAW, the 5D3 isn't a 1000 line camera such as the C100/C300 (the FS700 is just under the C300: http://provideocoalition.com/awilt/story/quick_review_canon_c300_super35mm_lss_cine_camera/P4/).

For outdoors, the high dynamic range of 5D3 RAW and uncompressed frames is clearly superior to the FS700's lower bitrate highly compressed H.264.

IPB breaks up with fast motion- 'mush'; it has to in order to maintain the bitrate. However, during playback it's not visible to the average viewer. We see it paused in an NLE and it looks terrible. For low motion shots on the 5D3, most people can't see a difference between IPB and ALL-I (if it was easy to see, we wouldn't see the common question for which is better). When this thread was started, 5D3 ALL-I was worse than IPB (the images posted at the beginning of this thread show this to be true). Those issues have been fixed- the question now is whether ALL-I is better for a particular use vs IPB (disk space vs. any visible improvement for the types of shots for the project: a 500GB IPB shoot becomes ~1.5TB if shot with ALL-I).
 
Celluloid Dream said:
(...) Avoid 5DtoRGB, MPEG Streamclip, ffmpeg, ffmbc, Prores, and the free AVID Quicktime verison of DNxHD if you use Windows.
Note: This does not include the new DNxHD codec in mxf format that comes with Adobe CC. This one does not use Quicktime and it is one of the best codecs I have ever tested in Windows. The 10-bit 4:2:2 175x variant is basically indistinguishable from uncompressed 10-bit 4:22.
Click to expand...
Hi Celluloid
Would you care to expand on this? Like so many others, I am trying to find a workflow for Magic Lantern RAW video that strikes a good balance between quality and efficiency (both temporal and spacial :)) and I was under the impression that DnxHD was a good alternative to proress for intermediate files and masters? Not sure exactly WHEN you're are suggesting to avoid quicktime stuff.
I have gotten a habit of using h264 in a quicktime-container for delivery and dnxhd for intermediate on my CS6 Win8x64box
 
You must log in or register to reply here.
Back
Top