It would be nice to have the option to process RAW data as B&W and use all the available photosites for higher resolution, or am I just dreaming in B & W?
[a href=\"index.php?act=findpost&pid=97116\"][{POST_SNAPBACK}][/a]
If nothing is done to interpolate the 2 missing colors at each photosite, then what the RAW data is, is a potential checkerboard pattern depending on the color in a given area. Neighboring pixels are only equally sensitive for most digitals when the combination of subject color and lighting has a magenta cast; 2x as much red light as green, and blue somewhere in between. If you try to accomplish this by multiplying the red-filtered pixels by 2, and the blue-filtered pixels by 1.4, for example, the sensitivity will be equal for white and grey subjects in daylight, but you will still get a checkerboard pattern when there is a moderate degree of saturation in the subjects.
A good demosaicing actually increases luminance resolution; that's the main point of it (as opposed to merely interpolating the missing colors at each pixel from the neighbors). If your subject is bright red, then only the red-filtered pixels are going to have much overt luminance information. Demosaicing tries to use the trends in the blue and green pixels, which are slightly sensitive to red, to fill in the luminance detail.
IMO, a good B&W RAW converter would have an overlay of a high-pass filtered version of a luminance-oriented demosaicing, to apply to something like a channel mixer that worked directly on the RAW color channels (without any of the hue shifts or color-selective saturation boosts performed in typical conversions). Or, perhaps a channel mixer with 6 sliders instead of 3; one for high-pass and one for low-pass for each channel. That way, you can get the general tonal curves from just the red channel if you want, but also get some pixel-level luminance detail from the channels you don't want to affect the tonal curve.
That said, I don't think a typical color conversion is the best place to start for a B&W; a conversion that didn't alter saturation and hue from the RAW data would be the best choice; then what you get is more like what the camera actually captures in the RAW data; three interleaved B&W images with three different color filters. A good B&W RAW converter is needed, IMO. Color-oriented converters are not doing as good a job as a dedicated one could do.
