Making Foveon-like Images with a Bayer Camera

NordicFlyer wrote:

One more try...this time with Raw Therapee-Gimp methodology; final edit in CaptureOne 22. Linear response curve, Auto adjust exposure, contrast & brightness, HDR, and levels. White balance sampled in the same area of both images.

Very little difference between the two images outside of lightness and subtle colour differences. Which image is best? Are they "Foveon" like? You decide.

Capture One 22 NEF to JPG

Raw Therapy/Gimp methodology. Tif to JPG.

When I reduce the first one to 1/4 original pixels (2144-x1424) by the Nearest Neighbor method, it very closely matches the second one (2156x1430) at pixel level.

In this news item:

Sigma releases update on the development of its three-layer full-frame Foveon image sensor: Digital Photography Review (dpreview.com)

I can't link you to what follows because of the way the DPR news item comments are listed out, but Roland Carlsson writes:

"The problem is: it is NOT easy. It is not easy at all. There are at least six problems.

1. Controlling the 3D doping is a nightmare.
2. The upper layers mask the lower layers. Hard to get a balance regarding how thin the layers shall be.
3. There are lots of electrical circuit on top of the imager, blocking light.
4. The layers have very similar spectral response, making the image very noisy when increasing the saturation to a normal picture.
5. It is NOT RGB. Not even close. You need to convert it to RGB, but this conversion is ambiguous. You can get strange colors.
6. It is all really experimental."

A couple comments later, he writes about downsizing Bayer to 1/4 original pixels:

"It can be done and it has been done for APS-C, both real Foveon and the Quattro Foveon lookalike thing. We who have followed it know its advantages and problems. The reason for Foveon is that Bayer sux. It really does. It is very difficult to get clean results at pixel level. Window edges are sharp and foliage not. Colored aliasing might turn up where you do not want it. In my experience, to get really sharp and nice results you shall downsize it to 50%, i.e. 1/4 of the area and number of pixels. The problem for Foveon is that it is much easier to make a 80 MP FF Bayer than a 20 MP FF Foveon. So, Bayer does stay ahead of Foveon all the time."

Note: I corrected a few typos after pasting this text above.

To find these posts you have to sift through all 381 comments below the news item, unfortunately.

Roland Karlsson: Member profile: Digital Photography Review (dpreview.com)

These quotes are listed in his recent posts.