BEST settings for sharpening to prevent Aliasing in Adobe Camera RAW

[fotosav]

Hi!
First, many thanks for ML! RAW video is amazing!

I using Canon 5D mark III and MLV module (2014Feb11, 2014Feb24 builds).
This is my first experience and I noticed the raw video frame (.DNG) and sRAW still image (.CR2) are different.
The DNG has some aliasing artifacts and more noise - see picture.
I have tested scaled and crop modes, the result is RAW video has aliasing artifacts and more noise for the same Camera RAW settings.

Questions:
1) Is it normal?
2) Have others tools the same aliasing for output? Can anyone convert this DNG-frame use DaVinci Resolve? Resolve Lite is not working on my computer.
3) What is the optimal settings for Sharpening and Noise reduction in Adobe Camera RAW for RAW video?
Please tell about your experience.

DNG frame:
http://fotosav.ru/temp2/sample.zip


I use Adobe Camera RAW 8.3:
Sharpening = 40/1.0/25/0 (Sharped)
Noise = No reduction (Default)

===
Updated.
Resolve vs Adobe CameraRaw:

[Midphase]

Well, understand that while CR2 stills take advantage of the entire sensor resolution, while raw video is still using binned pixels to arrive at the 1920x1080 resolution. So yes, some aliasing, moire and softening are inevitable.

It would be insanely incredible if the cameras could somehow record the full resolution of the sensor 24 times per second, but it's not possible and hence the degradation in the quality as compared to still raw images.

[fotosav]

raw video is still using binned pixels

What "binned pixels" mean? I use crop mode - what is difference between CR2 stills and DNG raw? As I understand the crop mode get 1:1 actual data from sensor like CR2 - is it correct?

Thanks anyway for answer!

[SteveScout]

But how did you manage to get exactly the same framing from the H.264 and the crop mode? You would need to use another lense to correct for the crop mode .. much much wider. From the picture I´d assume that this is not crop mode .. and this is why you see binning artefacts.

[fotosav]

But how did you manage to get exactly the same framing from the H.264 and the crop mode? You would need to use another lense to correct for the crop mode .. much much wider.

Yes, exactly - I use f=50 mm for H.264/photo shots and f=16 mm for crop video (16*3=48mm). See depth of field. The crop mode has wider DOF (because f=16 mm).

From the picture I´d assume that this is not crop mode .. and this is why you see binning artefacts.

It is crop mode. The crop mode and not crop mode have the same artefacts.
Is there information about how ML receives data from sensor and converts it to DNG?

[Midphase]

Hmmm...in that case maybe this has more to do with the way the image is being debayered?

I honestly don't know why in crop mode you're getting different results than in still mode. Perhaps one of the developers could chime in with some insider info?

[ted ramasola]

what resolution in crop mode did you use to record that image which showed aliasing?

[fotosav]

what resolution in crop mode did you use to record that image which showed aliasing?

Res: 1920x1080 | Camera: Canon EOS 5D Mark III | Lens: EF16-35mm f/2.8L II USM at 16 mm (79 mm) f/8,00 | Expo: ISO 320 1/33 s | Style: Landscape | WB: 5200K

ted,
do you use Da Vinci Resolve Lite? Could you convert my DNG (http://fotosav.ru/temp2/sample.zip) to TIFF via Resolve? Unfortunately, I cannot run Resolve on my computer.

[mageye]

In the images that you used for examples I can see that there is a depth of field change. If you look at the scratches on the metal that appear in the DNG file they are out of focus on the others. So, for a start you are not comparing the same images.

I just did a test where I set the camera up on a tripod and took a full resolution raw file (CR2) using a infra red remote trigger (to not move the camera at all). I then went into live view and took a 3x zoom (2144x912) MLV RAW DNG sequence.

I then took them into my computer and extracted the DNG sequence and loaded up the full resolution CR2 image into After Effects. It's a large frame (5616x3744) I then loaded up the DNG sequence and changed the opacity of the layer so I could align them properly. I used exactly the same ACR settings for both.

I could then turn the layer visibility on and off to make a comparison.

The images aligned perfectly. It is indeed a crop of the bigger image. I noticed a very slight line down the left hand side of the cropped frame. It's nothing that would be a problem because it's on the edge of the frame and it's not fully black. Just very slight.

There is also a very slight difference in the rendering of the two images. The CR2 seems to contain more grain. Again it's not that much of a deal (at least I don't think so). The DNG is fractionally softer. If anything, that softness will be some kind of aliasing artefact and it does make it look smoother (or fractionally softer). None of it is enough to make me shriek in horror though.

I am not really getting any nasties there. I suggest anyone that is concerned about the difference should try it and see if it is of enough concern to be a show stopper. It's not for me.

This is just my humble opinion of course.

EDIT: In fact. I am a muppet!!! . I forgot to fade up the opacity on the layer and slight misalignment would make it more blurry!!! Now I don't even see any difference in the grain levels! YAY! The only difference is a very slight shift in colour balance (again it's very slight and would be easily adjusted).

[fotosav]

Hmmm...in that case maybe this has more to do with the way the image is being debayered?

I have converted DNG via DaVinci_Resolve_Lite_10.1.2.
It looks like Resolve get more accurate result (and more soft) - Resolve vs ACR: