Incredible Artifacts disappear with Z1 Downconversion

[Peter Rixner]

Hi :)

I did a shoot, what might be the worst case for MPEG compression.
A very tele-shot of a running horse. Of course as I am constantly paning with it, the landscape is totally blurred and completely changing in a few frames.
OK. So I wasn't really surprised when I captured (with Premiere and the supplied Cineform) and the exportet uncompressed TGAs to Aftereffects and found horse and everything around it, looking like the worst JPEG if ever seen. Big parts of the horse where one brown color or only a few brown bands. The lanscape isn't much better.

So I thought. Well, you can't have everything, added some film-grain to cover the artifacts and was about to accept that.

Now in the End of this production the HDV images are scaled down SD for some reasons.

So I thought, maybe I give the in-camera downconversion another chance.

AND INCREDIBLE:

All MPEG artifacts are COMPLETELY gone. It looks like it alway has been a DV with perfect motion and wonderful colors and detail.

So ... what is telling us that ?

I knew that the in-camera downconversion is better when it comes to field-flickering compared with all other software I tried, but I never saw a difference in colors or let's say color-banding.

So the information seems to be on the tape. Where is it lost then ?
In the cineform codec ? Is it possible, that it's not as good as the "real" mpeg on tape ?

So no problem for this time but this must be solved for the day it's not going to sd in the end.

Thanks for every help or idea.

Peter

[Colvin Eccleston]

I have found that the captured image quality is affected by the capability of the computer. Building a more able workstation might help. I noticed some improvement just by swapping to a modern videocard and the effects were not just in monitoring but captured to disk.

[Steven White]

Would you mind posting an HDV frame with all the artifacts, and a DV frame from the in-camera down-conversion? If you don't have web-space, e-mail me and I'll host it.

Most of the blocking in footage I've shot is worth about 4-8 pixels in the HDV space. When converted down to DV, these artifacts are gone, simply due to the re-scale.

-Steve

[Peter Rixner]

Thanks so far. The solution is coming closer :)

Colvin:
I don't think the workstation is affecting it. It's captured via firewire digitally and then converted in that "intermediate" format of cineform. The Videocard does'n affect that. Maybe displaying it, but not when capturing.

Steven:
When posting the material I found, that much of the artifacts are hidden in interlacing and jump out quite clearly when deinterlacing.
Anyway they're gone or MUCH better when downconverted with the Z1.

The interlaced 1080 looses the interlacing. At least when downscaling in Aftereffects. And so if wanted deinterlacing or not it's progressive afterwards. And the artifacts are then seen clearly.

We had a discussion on that downscaling here some weeks ago. And as far I remember there was the common sense, that scaling down in Premiere and Aftereffects doesn't work as good as with the Z1 downconversion.
I wonder, what they are doing inside that black thing ... in realtime !!!

Anyway, here are the images:

http://www.perix.de/dvinfo/dvinfo.html

So ... further ideas or suggestions ?

Thanks!

Peter

[Ruslan Odintsov]

Originally Posted by Peter Rixner

I knew that the in-camera downconversion is better when it comes to field-flickering compared with all other software I tried, but I never saw a difference in colors or let's say color-banding.

Amen! I've always been telling people on this board that Z1/FX1's downconversion is superior in fields rendition than any software that's commonly available. Nobody believed me. So, thanks for posting this thread.

Also, check out Z1 BBC UK thread and see it for yourself on page 5 where there are posted M2V examples of Canopus vs. FX1 downconversion. They say that software downconversion was better, but to me FX1 was definitely way better on horizontal lines.

[Steven White]

Well, I tested a few theories... but I got nothing exactly.

The only way I get close is with a combination of scaling each field independently, blurring and applying a sharpening filter. However, the DV compression codec also adds some artifacts and kills color-space... even then, the banding patterns very different in the DV version - as if some adaptive blurring and sharpening process is used. (for example, I can nearly match the background and the small horse, but not the foreground horse - very aggressive blurring is needed to smear out the banding artifacts there, at which point, no amount of sharpening restores the detail on the small horse).

I bet there are some propriety algorithms similar to what's in their televisions at work here. There is a very significant decrease in sharpness however with the DV version - particularly noticable in the background. Furthermore, there could be some mal-treatment of the HDV or DV colour-space that's not being accounted for in software.

-Steve

[Ruslan Odintsov]

Originally Posted by Steven White

There is a very significant decrease in sharpness however with the DV version - particularly noticable in the background. Furthermore, there could be some mal-treatment of the HDV or DV colour-space that's not being accounted for in software.

Perhaps that is somewhat happening with DV codec, but if you connect Z1 through component connection and set it to downconvert 480p/576p output, you get a perfect, smooth and very sharp picture at 60/50 (or i guess 59.9? for NTSC) full frames per second. No artifacts, no horizontal lines issues at all. Even with 480i/576i donwconversion output through component it looks great. It easily beats any software downconversion. It looks just as good as the best progressive 24p DVD's, with one significant advantage: smooth motion because of increased frame rate. So now the only question is: how to capture this component signal without shelling around $2k for a professional capture card? :(

[Steven White]

It easily beats any software downconversion.

This is a fallacy. Software down-conversion can be done any way - and since both the camera and the software are working from the original MPEG-2 transport stream, whatever the camera does (save analog filtering) can be exactly reproduced digitally without any Digital-to-Analog-to-Digital losses. It's only a matter of knowing the algorithm.

-Steve

[Peter Rixner]

All of this is really interesting. At the moment I am happy again to know, that it is possible to get a good picture. Yesterday I really doubted that whole HDV.

And again: I think that's the worst case for MPEG. Rapidly changing images.

OK ... I hear my friend Thomas say: "It should be able to record a running horse ....."

Everything else I've shot throught the last month had no noticable artefacts.

So far ... even this discussion almost solved the problem.
Still - if anyone knows more solutions ... go ahead :)

Thanks!

Peter

Uh ... by the way ... did you notice, that in this HDV days, there is no flickering in the underscan areas ? Thats really a fine thing. I always had to crop edges in former times - which was 4 month ago :)

[Steven White]

IMO the "worst case" with HDV is to simply give on the MPEG-2 and tape DV straight up. But like this test illustrates, the in-camera conversion is good enough that it's not even worth doing so.

-Steve

[Richard Entwistle]

Peter,

Is the Z1 Steadyshot affecting your pictures?

Does anyone know how the Steadyshot ON when using a tripod behaves? As per PD150/170, SS should be off for tripod shots. Maybe if on, the Z1 could 'blur' more in HDV.

I collect a Z1 next month and looking forward to clear DV widescreen too. :)

[Ruslan Odintsov]

Originally Posted by Steven White

This is a fallacy. Software down-conversion can be done any way - and since both the camera and the software are working from the original MPEG-2 transport stream, whatever the camera does (save analog filtering) can be exactly reproduced digitally without any Digital-to-Analog-to-Digital losses. It's only a matter of knowing the algorithm.

Steven, it's fallacy to think that a dedicated hardware solution is not good enough. There is nothing, I repeat, nothing that beats HDV shot on Z1 to PAL/NTSC downconversion through component output of Z1. If you cannot capture component output then the next best thing would be in-camera DV downconversion. You can always use plug-ins to filter DV color deficiencies. I'm sorry, but none of the commonly used software have any advantages over in-camera downconversion except one: it's easier to achieve better color reproduction because you go from MPEG2 to lossless/uncompressed. However, every other aspect of software downconversion is tremendously inferior to in-camera downconversion. Those horisontal flickering lines are absolutely awful (again look in BBC UK thread with Canopus samples link).

Those people who are using software downconversion solutions are in denial. The colors are better so these people lie to themselves (and everybody else) that the overall picture is better as well. Unfortunately it is not. Just look at the Canopus samples and see for yourself. If one doesn't see the inferiority of the software solution, then there's something wrong with one's eyes.

As for knowing the algorithm for natural looking downconversion, well, it appears that none of the software manufacturers (including Sony with their Vegas) know it, or want to include into their products. So I guess somebody just needs to hack into the algorithm of Z1 and write the software code for it, if it's at all possible.

[Peter Rixner]

Richard:

the steadishot was off, when shooting that pictures.
But I really don't see much problems when it is on. Could be even an advantage if your tripod's paning isn't the best.

When I mount the Z1 on Steadicam I use steadishot as well.
Never had a Problem with it.

Peter

[Gino Do]

Have you tried using CAPDVHS application? and then convert to the uncompressed format. Cineform is an intermediary Codec so it does some converting and manupilation of the video, on top of the original on camera compression. I think CAPDVHS just captures the original compression.

I've been shooting HDV and had the same problem with the artifact problem using cineform/premiere capture work flow. The artifacts went away when I change my work flow to CAPDVHS and up convert the footage to the targa uncompress format.

Hope this helps.

[Kurth Bousman]

one question amigos - I'm not at home or I'd try ( maybe next week ) but can you downconvert and record component to a dvcam deck like dsr20/40 ? or do you need a $2k video card ? thanks