Image degradation with Cineform codec at Filmscan2?

[Marc Syp]

Hi guys.

I am new to the HD encoding world so forgive me in advance if I'm missing something obvious.

Recently I did a project at 1280x720p (30fps) that required a smoothing algorithm mid-way through production. My original footage was original Quicktime .mov footage from a Panasonic FZ-28 dropped straight into AE CS5, which was then matchmoved with a rendered sequence at 1280x720p.

After the matchmove and rendering had already been done, I discovered the need for a smoothing algorithm (VirtualDub + Deshaker). After some research, I found that people recommended using the Cineform intermediary codec for round-tripping. I did a number of tests and was a bit nervous about using cineform based on the results.

I exported three different AVIs from my AE timeline. Cinform "High", Cineform "Filmscan2", and AVI uncompressed. Attached you will find still frames from each of these sequences (I loaded these clips into AE and grabbed the same frame of each).



Filmscan2

http://www.mediafire.com/imageview.p...siu66y&thumb=6


Uncompressed

http://www.mediafire.com/imageview.p...cfcovh&thumb=6


I recommend downloading for closer inspection...



To me, when you blow these up, there is a clear difference between the uncompressed and the cineform AVIs, which have significant blurring and some noise artifacts especially at contrasty edges. Based on these tests, I decided to use the uncompressed AVI for roundtripping to VirtualDub.

However, on the return trip I tried both uncompressed and cineform Filmscan 2 and in the final encoding pass (h.264 mp4 4Mbps and WMV9 4Mbps) I was unable to tell the difference.

Is there something about the intermediary codec that causes it to be blurry and/or noisy during editing but that reconstitutes itself in final encoding? Or do I have some settings wrong? Does it not work as well on compositions, is it only meant for raw footage?

I would really like to use the cineform codec if it is capable of "visually lossless" encoding. (I can't afford to give up 100s of GBs to uncompressed video...) Note, in the future I will start using a T2i or 60D and will probably transcode the camera footage immediately before editing.

Any help/tips are appreciated.



Thanks,
Marc

[David Newman]

Nothing going wrong with the compression, it is working perfectly. All you are seeing in the difference between 4:2:2 and 4:4:4, if you used 4:2:2 uncompressed it would have the same artifacts. You need to select 4:4:4 encoding to match a 4:4:4 source (the chroma resolution of the overlay.) However, your target of H.264 is 4:2:0, so it will lose more chroma information than 4:2:2, and as a result there is no advantage in using 4:4:4 or uncompressed.

Here is the same image encoded at FS2 as 4:4:4.
http://cineform.com/temp/FS2-444.png

No visible difference.