Aliasing with RAW

Started by Viente, June 07, 2013, 02:08:29 AM

Previous topic - Next topic

0 Members and 1 Guest are viewing this topic.

Viente

I'm sorry for my ignorance but why 5Dm3 doesn't  have any  aliasing in RAW video, but all other cameras do?

Thanks

1%

I bet it does... they just tried real hard to reduce it.

togg

Quote from: 1% on June 07, 2013, 03:44:02 AM
I bet it does... they just tried real hard to reduce it.

I'm interested! How was this possible?

reddeercity

Hello All  :)
With the 5D2
I would say that Aliasing in raw is minimal Compared to H264,
i thinks because the image is Sharper and cleaner
you ten to focus better & you can get away with a soft focus
where you can sharping later to  minimize Aliasing.

So i was interested to see how good the Raw DNG image is, in term of scalable.
So i took 1880x854 24p DNG file import to photoshop and upscaded to 4096 x 1861
maintained the 240 dpi, 16 bit tiff, Basic White Balance & Color Correction


when you play the video in the small screen you will notice moire & aliasing
But thats only in the small player go to full screen and there is no
moire & aliasing.
The Original file on the image quality tab is the 4K mp4 i uploaded .

noisyboy

Quote from: reddeercity on June 07, 2013, 04:40:58 AM
Hello All  :)
With the 5D2
I would say that Aliasing in raw is minimal Compared to H264,
i thinks because the image is Sharper and cleaner
you ten to focus better & you can get away with a soft focus
where you can sharping later to  minimize Aliasing.

So i was interested to see how good the Raw DNG image is, in term of scalable.
So i took 1880x854 24p DNG file import to photoshop and upscaded to 4096 x 1861
maintained the 240 dpi, 16 bit tiff, Basic White Balance & Color Correction


when you play the video in the small screen you will notice moire & aliasing
But thats only in the small player go to full screen and there is no
moire & aliasing.
The Original file on the image quality tab is the 4K mp4 i uploaded .

I hate to say it dude but the moire and aliasing on that shot is very obvious to me. On the fence between the trees on the left, color patterns all over the distant tree and all the weird coloring on the darker of the two rooftops.

Sorry man - don't mean to be negative, just saying what I see.

Oedipax

The moiré on the mk2 is killing me. I'm having a blast playing around with it, but unless it miraculously gets improved in a future ML build (which I know is probably beyond the scope of what's technically possible), I'd have a hard time shooting anything serious on this hack with anything other than a mk3. I can see aliasing everywhere in a lot of my mk2 stuff. And I like things like deep-focus landscapes too.

a1ex

Aliasing can't be removed in post without breaking the Nyquist theorem ;)

(which is on the same level as breaking physics laws)

kgv5

Quote from: Oedipax on June 07, 2013, 07:48:58 AM
The moiré on the mk2 is killing me. I'm having a blast playing around with it, but unless it miraculously gets improved in a future ML build (which I know is probably beyond the scope of what's technically possible), I'd have a hard time shooting anything serious on this hack with anything other than a mk3. I can see aliasing everywhere in a lot of my mk2 stuff. And I like things like deep-focus landscapes too.

So you have to get VAF-5D2b filter, it seem it makes a great job in reducing aliasing and moire. I am thinking of buying one for 6D.
www.pilotmovies.pl   5D Mark III, 6D, 550D

DerekDock

Yeah I'm running into a bunch of Aliasing issues on the 60D but I think it only looks worse than normal h264 because the detail level is higher. Tomorrow I might be on the search for some post plugins to help reduce it but for now the trade off of shooting RAW is much better. Even my Big Sur video on the MarkIII had some aliasing on the opening bridge shot but then I think that's cause the sharpness was set to 25 in Lightroom.
www.vimeo.com/leftcoastdigital
www.twitter.com/derekdock

Kraemer

I don't quite know whats going on with that last shot reddercity ???, but here's my first ML raw test: 
The Aliasing and Moire are GONE!

https://vimeo.com/67864808

This was shot on the 5D3, handheld at 1920x1080. It's 6 sec long.  The RAW image is stunning.  The RAW codec, (or lack of) has eliminated
most, if not all of the aliasing and moire present in the building.  The first 3 secs are H264, the last 3 are RAW.  Notice the
windows.  RAW took all the false coloring out of the windows and looks almost futuristic.  On the side of the bldg, aliasing was going on, and with the RAW-- it's not even there.  Just amazing.  The latitude, sharpness-- Thanks Magic Lantern Team!!  Can't wait to learn the 3x crop and make something better than a test!

Nachelsoul

I think the secret to avoid it would be while shooting. I got better results with the 5D2 in terms of aliasing and moire using a ND (0.9) and a Pola together in camera, and avoid super sharp lenses like Zeiss and Canon EF. A very good choice are the Leicas R, quite sharp but also soft/creamy, also vintage lenses not sharp like Meyer's and Jena east German ones. Also notice to get nice results using wide apertures, a little bit soft but less noticiable aliasing, cause it is still there  :-[

DerekDock


Quote from: Nachelsoul on June 07, 2013, 09:23:19 AM
I think the secret to avoid it would be while shooting. I got better results with the 5D2 in terms of aliasing and moire using a ND (0.9) and a Pola together in camera, and avoid super sharp lenses like Zeiss and Canon EF. A very good choice are the Leicas R, quite sharp but also soft/creamy, also vintage lenses not sharp like Meyer's and Jena east German ones. Also notice to get nice results using wide apertures, a little bit soft but less noticiable aliasing, cause it is still there  :-[

But what about all the money I invested in amazing Zeiss glass :P
www.vimeo.com/leftcoastdigital
www.twitter.com/derekdock

Nachelsoul

Quote from: DerekDock on June 07, 2013, 09:42:10 AM
But what about all the money I invested in amazing Zeiss glass :P
Zeiss is the lens to go in 5D3, sure!!!!. But with the 5D2 will be worth spend some money in the Mosaic Aliasing filter. I love Zeiss and never get rid of my set, NEVER. These lenses will survive me.

apefos

I found a good solution, please take a look, try it and post your feedbacks:

http://www.magiclantern.fm/forum/index.php?topic=6248.0

reddeercity

Quote from: noisyboy on June 07, 2013, 06:04:23 AM
I hate to say it dude but the moire and aliasing on that shot is very obvious to me. On the fence between the trees on the left, color patterns all over the distant tree and all the weird coloring on the darker of the two rooftops.

Sorry man - don't mean to be negative, just saying what I see.
You are over analyzing the image , you must take this
with a grain of salt. I did not take the time set up the shot
As I would in a paying Job.  That being said I used my
cheapest lens , sigma 50mm f1.4 which is very soft lens 
When Wide open. The Sigma dose not compare to
Canon "L" lens. And also if you would have read what I
said on the YouTube page I would only
use it as "B-roll" in a 4k project, if I new i Was going to use a 4K upscalded
images then I would shoot in a way to keep it clean.
Beside that I don't think YouTube dose a good job of coverting video content.
On my grading monitor i don't see the problem  you do, some monitor  are
not able to reproduce a clean sharp image from 4K
I think asus are the worst ones , I have alway had problem with  them, can't
tust the image for 1 second .

kgv5

I was wondering what gain of the picture quality could be achieved with 6D by using VAF-6D for raw video.
Many examples and comparisons i believe are made with some standard picture profile and increased digital sharpening to exagerate aliasing and moire to be much worse in comparison with filtered image.In real life most of us using cinestyle or similar picture style with sharpness turned all the way down to make aliasing/moire less visible. But...RAW is not using any camera settings, IMO thats why picture seems to have more aliasing and moire, especially power lines looks ugly as hell.
Could someone make such a comparison: raw original/raw with VAF filter? I can be on 5d2, probably will be the same effect as on 6D. I think there was one on YT but i beleive it was shot indoors and don't show all the effect.
www.pilotmovies.pl   5D Mark III, 6D, 550D

togg

Quote from: kgv5 on June 08, 2013, 01:56:17 PM
Could someone make such a comparison: raw original/raw with VAF filter? I can be on 5d2, probably will be the same effect as on 6D. I think there was one on YT but i beleive it was shot indoors and don't show all the effect.

I'm interested in this as well. It's not a small purchase to make.

Redrocks

I read somewhere on the site in the past few days that VAF filters are designed for the stock h.264 codec and that they are not suited to RAW. Would be nice to see comparisons on any camera though. 

a1ex

QuoteI read somewhere on the site in the past few days that VAF filters are designed for the stock h.264 codec and that they are not suited to RAW.

LOL. They are designed for 1080p video sampled from the entire sensor; they may be too weak for 720p and they are way too strong for crop mode video.

kgv5

Quote from: Redrocks on June 09, 2013, 02:59:40 AM
I read somewhere on the site in the past few days that VAF filters are designed for the stock h.264 codec and that they are not suited to RAW. Would be nice to see comparisons on any camera though.
Well, its an optical element, not a software solution,I think it should solve most of moire and aliasing problems before picture is  written to card so I have high hopes for this. I have ordered one, so we will see in a couple of weeks.
Quote from: a1ex on June 09, 2013, 08:53:33 AM
LOL. They are designed for 1080p video sampled from the entire sensor; they may be too weak for 720p and they are way too strong for crop mode video.
n
Maybe, but isn't a crop mode made out of 1920x1080? (I mean picture with 1x magnification, resolution say 1792 x something)
www.pilotmovies.pl   5D Mark III, 6D, 550D

a1ex

In 5x it records at the photo mode resolution of the sensor (1:1 crop). In 1x, the image is downsampled by 3, and the filter is designed for this downsampling factor. In 720p, the image is downsampled by 5 vertically and 3 horizontally, and that's why the filter might be a bit weak here.

Roman

Quote from: kgv5 on June 09, 2013, 09:04:14 AM

Maybe, but isn't a crop mode made out of 1920x1080? (I mean picture with 1x magnification, resolution say 1792 x something)

Think of it as, because the sensor is downsampling in a way that means certain parts of the sensor are skipped,  there is a gap of a certain distance between parts of the sensor it records from.

A blurs the image an exact minimal amount, so the light that would have fell apon 'the gap' reaches the parts that are recording.

When you change how much of the sensor you are using, or using the full sensor at a different resolution, the size of the gap changes so the VAF filter will either now make your image blurry (if now less of a gap) or not work as well as intended (if more of a gap)

For what we're doing with raw, generally filming at higher resolutions with the full sensor, I would expect a vaf filter to remove aliasing at the cost of the image being blurrier than you'd like.

kgv5

Quote from: a1ex on June 09, 2013, 09:17:11 AM
In 1x, the image is downsampled by 3, and the filter is designed for this downsampling factor.

That's the most important for me  :)
I have found some very good examples VAF+RAW (the first one showing how much worse aliasing and moire is in RAW footage vs h264 - raw has always it's default sharpness)
http://www.youtube.com/watch?v=eeA-siZRE3A
http://www.youtube.com/watch?v=99ne_fDZFGM
www.pilotmovies.pl   5D Mark III, 6D, 550D

JohnBarlow

One of the image artifacts lumped under the heading moire/aliasing is the annoying color rainbowing that you see.

Avisynth has a filter which works wonders with this.

Roman

With the 5D3, is it possible that they are sampling from the entire sensor, rather than skipping?

So for example to make one pixel, they take an average from an entire area rather than skipping sections and reading from only one part.