|
Re: And mathematics tells...
In reply to Joseph S Wisniewski,
5 months ago
|
Joseph S Wisniewski wrote:
Emacs23 wrote:
mutatron wrote:
I could be wrong, but I see x-trans as the start of the post-Bayer era. It's only a matter of mathematics to get the image out of the sensor. Once that is solved and there are no impediments to using x-trans, more camera makers will start experimenting with different sensor arrangements, as a way of differentiating product and making people think their existing equipment is obsolete.
And mathematics tells since xtrans has lower resolution of red and blue (less samplings), it has lower resolution in general than bayer sensor.
You need some better mathematics. In general, the very highest resolution would be a sensor that skipped the red and blue totally. That's analogous to the human visual process, incidentally, except that we're a much more extreme case, with approximately a 20:1 ratio of luminance resolution to chroma resolution.
The xtans decreases NR problem complexity at the cost of demosaic complexity.
Its main advantages are in resolution: less obvious chroma aliasing, so you need a less aggressive antialiasing filter, and resolution approaches that of a monochrome sensor.
Efficient modern NR algorithms, on the other hand, matches high iso performance of bayer sensor to that of xtrans (assuming bayer sensor uses Sony IMX071, which is also used by fuji and forget about ISO6400 of fuji — it's close to 3200 with other cameras). From the point of IQ it's pure marketing trick, not actual achievement.
Odd then, that Dr. Bryce Bayer did his later work in pseudorandom patterns, such as teh xtrans, isn't it?
So, it's definitely looks like a brainless niche solution with no future. The future is with mutlilayer sensor with close to zero light loss, not that mosaic crap.
Interesting, considering that the only multilayer sensor currently on the market has approximately six times the light loss of a Bayer sensor. Can you tell us how a "close to zero light loss" multilayer sensor might be achieved?
--
Rahon Klavanian 1912-2008. Armenian genocide survivor, amazing cook, scrabble master, and loving grandmother. You will be missed. Ciao! Joseph www.swissarmyfork.com
Wrong. You don't know this field, sorry. I do.
Higher resolution of one channel with lower of other ones will lead to color artifacts at contrast places (higher resolution of the channel means it can have variation that isn't matched by other color sources). Thus, it needs to be smoothed to match them. So the resolution is limited by worst resolved channel. Sad truth.
Well, suppose Fuji has managed to really dig something that solves the problem (which is impossible, of course — lost information is lost information), will see at silky output vs other APS-C sized sensors:
NEX-7 Pentax K-5IIs
NEX-5n Fuji X-Pro1
Scale
Fabric
So, it's easy to see how better bayer pattern is. Just no contest. No wonder happened, the XTrans is inferior. I used only open source software here to generate jpegs from NEX-es and Pentax: Darktable with AMAZE demosaic algorithm and no sharpening applied, then did some deconvolution sharpening in GIMP.
| Post (hide subjects) | Posted by | When | |
|---|---|---|---|
| 5 months ago | |||
| 5 months ago | 1 | ||
| 5 months ago | 11 | ||
| 5 months ago | 1 | ||
| 5 months ago | 9 | ||
| 5 months ago | |||
| 5 months ago | 4 | ||
| 5 months ago | 4 | ||
| 5 months ago | 8 | ||
| 5 months ago | 2 | ||
| 5 months ago | 5 | ||
| 5 months ago | 1 | ||
| 5 months ago | |||
| 5 months ago | 2 | ||
| 5 months ago | |||
| 5 months ago | 3 | ||
| 5 months ago | 2 | ||
| 5 months ago | |||
| 5 months ago | 1 | ||
| 5 months ago | 4 | ||
| 5 months ago | |||
| 5 months ago | 1 | ||
| 5 months ago | 2 | ||
| 5 months ago | 6 | ||
| 5 months ago | 2 | ||
| 5 months ago | 1 | ||
| 5 months ago | 2 | ||
| 5 months ago | 3 | ||
| 5 months ago | 3 | ||
| 5 months ago | 6 | ||
| 5 months ago | 1 | ||
| 5 months ago | 2 | ||
| 5 months ago | 1 | ||
| 5 months ago | |||
| 5 months ago | 1 | ||
| 5 months ago | 2 | ||
| 5 months ago | 1 | ||
| 5 months ago | 1 | ||
| 5 months ago | 1 | ||
| 5 months ago | 1 |