HBR-IQ comparsions of various settings (25p or 30p)

towi

This thread is dedicated to evaluate the capabilities of different settings in HBR mode (25p or 30p).

I'll start with this: HBR 25p - Quantum X V2 rocket vs. Flowmotion vs. Sanity vs. Quantum X V2 rocket.

tungsten (lowlight), 180MB: http://wtrns.fr/__Oa704p1LJRpgM

daylight, 180MB: http://wtrns.fr/WF2e4oVg21Rjv3Q

(over the course of the daylight shots the light has changed, however the comparision is not entirely useless).

Comparison of 24H vs. HBR of the respective patches.

Quantum X V2 24H vs. HBR 25p - Flowmotion 24H vs. HBR 25p - Sanity 24H vs. HBR 25p.

For these I first transcoded the MTS files to QT MOVs (ProRes422 HQ) through 5DtoRGB at 25p

daylight, 180MB: http://wtrns.fr/KoZfc797Jc5nix

tungsten, 180MB: http://wtrns.fr/JZj4cqWg1AgKq-f

For reference, here's also a comparision at 24H of the tungsten shot, 115MB: http://wtrns.fr/BvJYfswBCrXgQM

Downloads expire March 10th.

Movies use Avid DNxHD codecs. You can get them here (for free, of course): http://avid.custkb.com/avid/app/selfservice/search.jsp?DocId=263545

Settings used on all shots:

smooth, -2|-2|-1|-2. ISO200

Care has been taken to avoid the ISO bug on all shots.

Lens used on the tungsten shot: Nokton 0.95/25

Lens used on the daylight shot: Zeiss 2.0/50

My preliminary conclusions (based on these static scenes):

Quantum X V2 rocket:

great at 24H
great at HBR when the codec isn't pushed to its limits
compressed/noisy at HBR when the codec is stressed much (I am sure Driftwood will sort this out sooner or later… perhaps even in an update of QXV3 rocket later today)

Flowmotion:

great at 24H
HBR looks mushy in both the tungsten and the daylight shots

Sanity:

great at 24H. It shows less macro blocking than QV2 rocket and FloMo in the tungsten shot
great at HBR also when the codec is pushed to its limits

Have fun!

Kihlian

Great work thank to post thys, I work with HBR and that was just what I needed to get more clarity around all thys new settings! P.S. what you thynk about cake?

towi

what and where is "cake"?

Kihlian

I mean cake settings from Balazer. See: GH2 constant quality variable bit rate encoder settings, topic.

LPowell

@towi - "For these I first transcoded the MTS files to QT MOVs (ProRes422 HQ) through 5DtoRGB at 25p"

While I appreciate your efforts to make fair comparisons, I have to object to your labeling your results as :"Flow Motion" when they have been adulterated by transcoding the files into 5DtoRGB. Once you do that, you are no longer examining the original video recorded by the camera, you are evaluating a 5DtoRGB encoding. My understanding is that 5DtoRGB advocates use it precisely because they prefer the way it filters the video to the original recording:

http://nofilmschool.com/2010/06/is-5dtorgb-the-dslr-post-production-solution-weve-been-waiting-for/

"The resulting files are the absolute highest quality you’ll ever get out of the camera. In fact, you could argue that they’re even better than the camera originals since they’ve undergone high quality chroma smoothing."

If you're going to make blanket subjective judgements like Flow Motion HBR :"looks mushy", I expect you to back this up with downloadable unadulterated MTS files that demonstrate what you're talking about.

Original footage or it didn't happen.

towi

@LPowell

only the 24H-HBR (so 24p + 25p) comparision was transcoded through 5DtoRGB (to transform 24p into 25p easily). All other comparisions are based on direct import of the MTS files into Avid - no prior conversion through 5DtoRGB.

Besides ... when I open the MTS files in VLC or in Quicktime (with Panasonic's QT plugin "AVCCAM Importer" to play MTS files natively) things look the same. With Flowmotion there is clearly a noticeable drop in IQ when you compare 24H and HBR (25p) ... idependant of the software I use to view or transcode the files.

LPowell

@towi - "only the 24H-HBR (so 24p + 25p) comparision was transcoded through 5DtoRGB"

For these I first transcoded the MTS files to QT MOVs (ProRes422 HQ) through 5DtoRGB at 25p

daylight, 180MB: http://wtrns.fr/KoZfc797Jc5nix

tungsten, 180MB: http://wtrns.fr/JZj4cqWg1AgKq-f

And this is what you based your one-word evaluation on?

towi

please read my post again (or not... that's up to you). For the HBR comparisions of the 3 hacks I did not use 5DtoRGB to convert the files.

"And this is what you based your one-word evaluation on?"

no. on these comparisions:

tungsten (lowlight), 180MB: http://wtrns.fr/__Oa704p1LJRpgM

daylight, 180MB: http://wtrns.fr/WF2e4oVg21Rjv3Q

LPowell

I downloaded those two zip files, which contain Quicktime mov files. However, all I see in the files are still images, there is no video motion to evaluate.

Is your intent to use these patches for still photography?

towi

@LPowell - my intend is to do fair comparisions... which is actually only possible with a fixed camera. of course such comparisions are limited to evaluate noise, macro blocking and maybe the "sharpness" of the shots. motion is in fact also a very important factor ... but not as easy to show in side by side comparisions.

But I am wondering... are you saying that with motion in the scene Flowmotion out of the blue switches on its turbo and finally delivers usable HBR footage? Sound good, I have to try it out...

@Kihlian - I've just tried "cake" with the codec stress chart in HBR mode (25p). Attached a screenshot form Streamparser.

LPowell

@towi - "my intend is to do fair comparisions... "

OK, but from what I've seen so far, you're shooting static still scenery, extracting a single frame of your choice, transcoding that into a different codec, and then issuing one-word judgements.

Sorry, but your methodology is not what I'd consider a fair comparison of motion picture quality.

Also, your leading question "...are you saying that with motion in the scene Flowmotion out of the blue switches on its turbo and finally delivers usable HBR footage?" In my view, this clearly indicates your bias and agenda, and gives me reason to doubt the impartiality of your selection of screen shots.

Again: Original footage or it didn't happen.

towi

"you're shooting static still scenery, extracting a single frame of your choice"

press the play button! of course these are not single frames.

"In my view, this clearly indicates your bias and agenda" Neither bias nor agenda. I really don't care about the name or the creator of a certain hack. I've tested different hacks and would have loved to use Flowmotion. But HBR mode simply doesn't cut it for me.

filesize of all the original footage is to big to upload. but I suppose to do your own comparisions in HBR mode of various hacks and share them here. And I also suppose this to anyone else. There is nothing to hide - all the hacks are available here on the forum. Load them and do your own comparsions. This would be helpful for the forum community. Initially exactly this was my intend to start this thread. Sorry you are pissed off. That was certainly NOT my intend. It just happend ...

LPowell

@towi - "filesize of all the original footage is to big to upload. but I suppose to do your own comparisions in HBR mode of various hacks and share them here."

OK, here's my original Flow Motion v1.11 footage, downloadable from Vimeo (which is significantly more detailed than Vimeo's website compression);

towi

also show me Quantum X and Sanity or antother hack of the same scene... we are talking about comparisions.

towi

@LPowell here are 2 MTS files of the tungstens shot; Flowmotion and Sanity: http://wtrns.fr/E_o74eGw1OjZi-5

LPowell

@towi Congratulations, that motionless, low-detail, softly-focused shot featuring large amounts underexposed dark areas required less than 24Mbps peak bitrate in Flow Motion (as there is no motion to encode!). For a static, largely underexposed shot such as this, Sanity's long-GOP encoding did indeed produce higher bitrates.

To my eyes, what this comparison demonstrates is that even in worst case scenarios, Flow Motion v1.11 delivered results that were visually indistinguishable from Sanity. If these files weren't clearly labeled, I would have been unable to confidently tell them apart (without resorting to technical analysis). Not even the subtle color banding at the bottom center of the frame shows any telltale differences in the two encodings.

towi

"To my eyes, what this comparison demonstrates is that even in worst case scenarios, Flow Motion v1.11 delivered results that were visually indistinguishable from Sanity. If these files weren't clearly labeled, I would have been unable to confidently tell them apart"

you don't see a difference when you look at the folding ruler?

"softly-focused shot"

A "softly-focused" shot that shows moiré :-)) …

Let's leave it at that: you don't see the difference and that's perfectly fine.

aljimenez

@towi I like your goals of giving to this community some comparisons. However, as @LPowell has articulated, you need to have conclusions that are justifiable by the data, else call them opinions or feelings and let the community see the comparisons and reach their conclusions. It's probably best to show your comparisons, which are very valuable for all of us, and perhaps give your opinion, but not claim it is necessarily fact from a few tests... Al

towi

@aljimenez - thanks!

"perhaps give your opinion, but not claim it is necessarily evident from a few tests"

actually I thought I did just that by saying "My preliminary conclusions (based on these static scenes)". But possibly it sounds different due to language barriers...

Howdy, Stranger!

Categories

Tags in Topic

Top Posters