Why STANDARD resolution has more resolution than FULL resolution?

[Miguel "Macgregor" De Olaso]

Doing some rendering tests in Redcine it is quite clear that rendering at "STANDARD RESOLUTION quality" (which i believe reads the image from a 2K debayering) outperforms the "FULL resolution quality" rendering in terms of resolution.

Even a 2K frame rendered in standard res shows a bit more detail than a 4K frame exported at FULL res.
Why is this happening?

Also, if you compare a 1K rendered file in REDCINE at FULL res against a 1K QT proxy file, again the dirty QT outperforms the resolution of the nicely and slowly exported file from REDCINE.

 

[Kenn Michael]

I've seen this as well Macgregor.

To my eyes the standard output is much sharper. BUT, in certain circumstances, in super high detail areas of an image there is slight aliasing that can occur in a Standard output that is smoothed out on a Full output.

However, I do like the added slight sharpness, so I do my outputs at Standard. The _H proxy is basically the same thing as Standard, however I do see a slight magenta shift in color as well as gamma with the QT proxies over a RedCine output.

 

[Miguel "Macgregor" De Olaso]

There is aliasing and there is much more noise issues. But i wanted to show the fact that it seems to be much more resolution there that we are not seeing.

 

[Patrick Tresch]

There is aliasing and there is much more noise issues. But i wanted to show the fact that it seems to be much more resolution there that we are not seeing.

I think Greame is so afraid of aliasing that he pushes R3d to the outmost limits of blur... I think that a tiny bit more aliasing could "improve" IMHO the "feeling" of resolution.

Pat

 

[Graeme Nattress]

The 2k rapid decode is quite aliasy - you're putting nearly 4k of resolution into a 2k frame without a proper downsample filter. It's like downsampling with "nearest neighbour". It cannot have MORE resolution than the 4k it comes from. Also, because the colour channels are treated as if they're co-sited, but they're not, that can "enhance" the appearance of resolution, but it's all fake aliasing. Look at a 4:4:4 RGB camera that uses pixel shift - you get the very same colour fringe type effects, and they're giving extra fake resolution there too.

The Bayer reconstruction algorithms cannot remove or reduce aliasing that exists in the raw data. That's why we use an OLPF.

Now, in Bayer reconstruction to RGB, you can indeed go for a "high resolution" approach. But you must also accept artifacts, like the infamous Bayer zipper artifact, and I'm not prepared to write code that produces that artifact. For one thing, it looks awful, and for another, as soon as you add sharpness or saturation the lack of co-sitedness across channels will produce nasty fringes, halos and artifacts. Again, I'm not prepared to do that.

I want smooth, continuous images that react well to all post production things you can throw at them.

Graeme

 

[laguun]

I want smooth, continuous images that react well to all post production things you can throw at them.

Graeme

I never saw footage which reacted well if you throw a 3D page roll at it in post

SCNR.

 

[Obin Olson]

rotflol 4k don´t you see the beauty in the famous 3d page peel??

graeme...glad your taking this approach...we need to have clean images with all post processes!