Bayer - Foveon comparison

Started Nov 23, 2008 | Discussions thread
jtra Contributing Member • Posts: 927
Re: Bayer - Foveon comparison

Arky wrote:

Feel free to download, modify, print, crop, re-post and point out the
obvious deficiencies of these comparison images.

So I did. I tried to match color profile of the Foveon image with the Canon image.

Here is original 50d small preview:

Here is original foveon small preview:

And here is my transformation of 50d image to "Foveon" color:

Larger version is here:
http://jtra.cz/tmp/50d-foveonized.jpg

The used transformation is here:

w=[0.0242979124172085, 0.974145663246175, 0.0829599914900766, -0.0347804059944083, 0.0975009361149797, -0.26305326156169, -0.0428469898944963, 0.0191707560142681, -0.00612339946564458, 0.0068507818266704, 0.975673496001818, 0.0188544963524508, -0.0276228859661246, 0.0154411314424653, -0.00229227685097862, -0.0526694996683543, 0.0193133372795335, -0.03332131594754, 0.039482660815525, 0.982617916402083, 0.0751145879377187]

newred = w[0] + oldred*w[1] + oldred*oldred*w[2] + oldgreen*w[3] + oldgreen*oldgreen*w[4] + oldblue*w[5] + oldblue*oldblue*w[6]

newgreen = w[7] + oldred*w[8] + oldred*oldred*w[9] + oldgreen*w[10] + oldgreen*oldgreen*w[11] + oldblue*w[12] + oldblue*oldblue*w[13]

newblue = w[14] + oldred*w[15] + oldred*oldred*w[16] + oldgreen*w[17] + oldgreen*oldgreen*w[18] + oldblue*w[19] + oldblue*oldblue*w[20]

where all colors are represented as real numbers in range 0.0 to 1.0. The weights were trained by a variant of genetic algorithm on those two images so when you apply them to other images, it may look strange (for example there are almost no green surfaces so mapping of green may be a way off).

The Foveon and foveonized 50d pictures look "same" enough for me regarding color.

Post (hide subjects) Posted by
Keyboard shortcuts:
FForum PPrevious NNext WNext unread UUpvote SSubscribe RReply QQuote BBookmark MMy threads
Color scheme? Blue / Yellow