Andrew Kolakowski wrote:Wayne Steven wrote:
People quote other's incomplete marketed works as complete too much, with no real experience in freshly designing something even as incomplete as their example themselves, let alone look into the dark and create something brand new.
Look what you said few posts above: "we need 8K to get perfect 4K"- it was your key argument for 8K sensor.[\quote]
I didn't say "we need 8K to get perfect 4K", where did I say that. Don't try to change what I say and try to put it in quotes as if I said it. How dare you Andrew. You have been deliberately trying to waste my time after I've been so nice to you carefully explaining things. If you want to be like that you can bugger off as well. It takes a bloody long time to write these things to sooth your ego. I'm just looking at your irrational stuff below, I think you should give up and stop trying to think others in a corner. Your arguments are not right and insulting, and apart from wrecking the thread I've had enough. Your temperament has been showing through your arguments for a while.
??????
For those that maybe a bit bamboozled by certain things here, I'll explain it clearly. Just remember the universe is often made up of simple things working together, making it appear complex. So, you have to read things, separating out the simple things and how they effect each other. In this case simple means work.
So 8k gives better 4k as you get a 4 pixel Bayer color group under a 4k pixel. It helps with Moir and aliasing, while delivering a naturally sharper image. But it does not get rid of things completely. The gap between the pixels still exists, but hopefully smaller than native 4k. So, there is still a gap around the border of the 4k pixel likely (some don't have much of a gap at all. I'm cutting this short), so a small anti-aliasing filter still would help, or 10k+ resolution, as per oversampling BM and Red already does (which I believe I might have mentioned before as better then 8k).
Now, extracting and calculating out 4k from 4k data, as a true 4k point cloud rather than secondary line test. It is as it seems a calculation if dubious quality, rather than perfect quality, as delayering also is. So, no guarantee of perfect pixels. Each pixel gives a little signals portion of what is under that point, usually mixed with surrounding pixels signal from low pass filtering. But it is something to work with. There are a number of elaborate techniques that can expand upon this data. But for 4k from 4k, you are deriving from neighbouring pixels, as in debayering, the likely hood of what was there in the scene. With no anti-aliasing filter you are getting more the pixel signal but still with some cross talk between the primaries, and a gap between pixels, and minor to deal with. But in terms of real definition, there is data to work with there.
Now, upscaling is not just as described. The industry is fairly ignorant (as most of US are to some naive extent). You ave to think, listen and learn to do more. To must ave the tools to think about how to do it better.
So, 8k does help Bayer, interpolation and filter, for lower resolutions. To get the real image you can blur afterwards (but frankly, the sharp Netflix stuff, like in the Expanse, looks good). It does help deliver most accuracy to the actual image than grainy s35 film, which is only and approximation. But if you look through 8k vision, you are going to see a lot of fuzziness too, but in image pickup the accuracy of the pixel helps the post workflow. Computational photography techniques depends on accuracy. Calculations. Sensors these days, not in some bygone era, offer 'up to' beyond old cinema film, and probably newer replacement stock. In the old days film looked so good through a lot of hard work to fake the image, by conforming the levels with filters, lenses, lighting, reflectors, different films stocks, make up etc, and heavy, costly post grading. But until electronic techniques, film grading wasn't too flash.
So, it really is about better tools for more work choices and better images to play with in post.