Going by this example, I would choose the software sample to work with. And the rocket look would not be difficult to match, should one wish.
|
|
Going by this example, I would choose the software sample to work with. And the rocket look would not be difficult to match, should one wish.
Bill, I'm not sure I understand.
I can match the Software one by blurring the Rocket. Blurring destroys fine detail. So how do I get it back ? Sharpening gets close, if you stand back a few feet. The real detail is diminished.
So if I have a certain $700 lens who isn't as sharp as another lens, by the same account I can just sharpen to make the two equal, in terms of sharpness ?
Can I sharpen a 1920x1080 image and make it 4k ? 3K ? 2K ? It's all a cheat.
Of course, using a Bayer sensor is a cheat too, but at least we can attempt to get the most detail possible from it, right ?
The Rocket version looks noisy, true. Don't fear the noise. It can be dealt with. Ever looked at a 4k film scan at 1:1 ? Even that has 'hidden' detail that can be extracted.
For that 4K experience. Not the 3K experience.
Software should always deliver a better quality than hardware when it comes to encoding, conversion and processing of video. I've had long discussions with Snell & Wilcox people over the years about this - software will usually take longer but will give the best results.
Scott
Scott,
I agree 100%. Software de-mosaic methods are very tricky these days, and make decisions based on edges and other almost 'fuzzy logic' methods that attempt to deal with the classic problems such as far away picket fences, mesh patterns, close up shots of bridal veils, etc. Throw in having it hold up in motion, and you have even more issues like crawling/roping etc. Hardware hates these types of algorithmic branches. The easy way out, just blur it and call it a day.
The other problem: patents and associated leaches :)
I, too, prefer the software rendered output. When you blur the Rocket output, it appears you desaturate the blue channel as well.
Less, Could you share these R3d's? I would like to try em out on my software/hardware combo, see if we get anything similar or some different results.
Reason is, I have never seen pixels sooo evident on our red rocket system.
Sure, I'll get some samples you can try. BTW these were high ISO, hence the noisy look.
I also have some r3d's that have been encoded into Cineform RAW that I will post.
That's right, cineform RAW .... preserving the raw aspects of the r3d that we all love so much, but allowing usage in other workflows.
Can you say 24 fps 4K playback full debayer / selectable debayer methods .... I knew you could! CPU only. No card. On the fly color changes.
( I am not selling or sharing this code, but will share some MOV's generated by it )
| « Previous Thread | Next Thread » |
| Tags for this Thread |