xionc
•

Forum Member
•
Posts: 97

Foveon X3 - resolution, DPI, interpolation, comparision with Bayer resolution - by Laurence Matson

Dec 25, 2012

8

"Every Foveon imager has x number of pixels stacked in groups of 3 ("triplets") distributed over a matrix with x/3 number of spatial locations. In the case of the SD1, SD1M, DP1M and DP2M cameras, there are 4704 columns and 3136 rows equaling 14,751,744 spatial locations, each of which has three pixels (R,G,B triplets) for a total of 44,255,232 effective pixels or, rounded up for marketing purposes, 45 MP.

The parity factor figured out ten years ago is that a Foveon imager is equal to a Bayer imager with about twice the number of pixels. In other words, the SD1-class imager with about 15 MP (not pixels but rather spatial locations) is at near parity with a Bayer imager with 30 MP (in this case, pixels and spatial locations) There are various ways to get to that number:

- Black&White resolution charts show equivalent line-pair results at about number. In other words, if an SD1 Foveon imager (with its 45 MP organized at 15 MP spatial locations) hits the number 32 on the chart, a Bayer imager with about 30 MP will as well. These are approximate numbers, of course, but at that level, a couple million MPs really do not make that much difference. (Using color charts, it is no contest, which may be one reason why dpreview, among others, have stopped using them.)

- A Foveon imager file enlarged to match a 30 MP Bayer imager will produce an image that matches it as well. In other words, uprezing an SD1-class file to 30 MP dimensions does not result in any great loss of acuity in a print.

- Reducing a 30 MP Bayer file to the dimensions of an SD1-class imager file also achieves an equivalent effect according to many, who love doing this kind of stuff.

DPI has virtually nothing to do with image quality within the parameters of the original image. It has a lot to do with image dimensions. For instance, the 180 dpi image only redefined to 360 dpi is exactly half the size of the original in terms of dimensions but exactly the same size in terms of information.

If you think of each Foveon imager spatial location as a full bucket of paint, you can more easily visualize the "less than twice as large in terms of information" concept, since the "paint" is thinned a bit during the interpolation (a polite word for "guesswork") process.

And again, with an open mind, you could now understand the "superiority" of the Foveon concept if only on this point alone. A Bayer imager typically has to throw out a pile information (as much as 75%) before recreating the matrix using Bryce's brilliant process. Although the end result looks great, each spatial location is only equivalent to a 25-33% full bucket of paint, on average; uprezing from there means beginning with less-than-full buckets of paint.

Finally, where dpi is somewhat-to-very important - depending on your pickiness - is with printers. All Epson printers, for instance, print using factors of 180 dpi. All Canon printers use 150 dpi factors. It has been argued by many for many years that you should always match your file to the printer's factor. I happen to think that is a good idea."

I hope the troll thread will be deleted and this will stay for those who UTFSE, having an appropriate topic.