On Mon, Sep 23, 2013 at 07:28:16PM -0700, Larry Colen wrote: > > We could also try going from the trichromat bayer filter to one > that has a finer resolution of color sensitivity, or even one that > has both "rods and cones", a site without a color filter that gives > raw luminance and no chroma information.
The first of those is the area that intrigues me most. Once two pixels have been reduced to the same tri-stimulus value, there's nothing that can be done to separate them again. But, as most photographers know, the colour sensitivity of cameras (be it a bayer matrix on a digital camera, assorted colour films, or even multiple monochrome images shot through different filters) doesn't match that of the human eye. I'm sure all of us have noticed a photograph where the colours don't match what we remember from the scene. The most common example is a flower blossom that appears to be a very different hue, but I've also seen the same thing in other areas (such as paint schemes of racecars). People who are seriously into the technical background of colour matching (including, but not limited to, photo printers) know how a perceived colour can change dramatically under a different light, which is why any colour standard also defines one or more illuminants. I'd love to be able to defer some of that decision making until later in the image creation process, and handle it in the post-processing. But that would require sampling the spectrum at more than three points. The CIE standards only tabulate spectra at something like 50 different values, so that should be sufficient - I'd think 16 might work well. One company (Fuji?) has taken a tentative step in this direction, and uses different shades of green in the two cells of their RGBG array. [They also introduced a 'rods and cones' type of sensor, where some of the photo sites measured highlight luminance, to gain an increase in dynamic range at the cost of spatial resolution]. -- PDML Pentax-Discuss Mail List [email protected] http://pdml.net/mailman/listinfo/pdml_pdml.net to UNSUBSCRIBE from the PDML, please visit the link directly above and follow the directions.
