On 07/11/2018 11:17 AM, Marcel Brouillet wrote: > I read about demosaic-ing and did my RAW-101 class :-). I'm still a noob on > this. I understand the process would involve a RAW sticher, the PTO being > generated on some initial set of (rough) converted files. Yet... > Vladimir, why do you have the need to separate the colors ? Can't the > stitching be done on the array of Bayer data ?
I was thinking about something similar to "Bayer drizzle" algorithm (soory, right now I can't find any good explanationon fof it, the information on google is rather fragmented) The input is bayer data, output is normal RGB image. > Care has to be taken that > red values be distorted and only put in red positions in the resulting > array, before mixed with other red values from other distorted arrays. But > then the resulting array is a giant RAW array, a panoramic RAW, that needs > to be de-mosaic'ed, white-balanced, noise-reduced, etc... The advantage > (quite certainly not worth the development, but theoretical) being that you > handle the RAW processing on the entire pano, afterwards. > > In other words, programmatic care has to be taken that « *The next problem > is that after remapping the Bayer pattern is destroyed* » be avoided. > > I foresee a possible objection, that raw values only have sense in relative > values within a single shot, and that two shots do not share the same > numeric conversion origins (i.e. measurement offset, inconsistent > calibration of the A-to-N converter). Is that so ? > I see no advantage in saving output as Bayer data. For example Rawtherapee handles tif files almost the same as raws. Vladimir > > Le lundi 9 juillet 2018 20:53:52 UTC+2, T. Modes a écrit : >> >> Am Montag, 9. Juli 2018 16:02:06 UTC+2 schrieb nadv...@suse.cz: >>> >>> It might be possible to convert the mosaic data to independent channels >>> with transparency: >>> >>> source: >>> >>> RGRGRG >>> GBGBGB >>> RGRGRG >>> >>> to: >>> >>> RTRTRT >>> TTTTTT >>> RTRTRT >>> >>> TGTGTG >>> GTGTGT >>> TGTGTG >>> >>> TTTTTT >>> TBTBTB >>> TTTTTT >>> >>> >>> I don't see any principal problem with processing such images with nona >>> and enblend. >>> >> >> >> If you do it this way each color will get different interpolation for each >> color (because different alpha channels) and this will results in color >> seams. >> The next problem is that after remapping the Bayer pattern is destroyed. >> With the destroyed Bayer pattern it because impossible (or at least very >> difficult) for the RAW converter to demosaic the data now to get the final >> image. >> >> (An alternative approach would be to demosaic the image data, then remap >> the data and then mosaic the data again to get the RAW again. But then you >> will loose resolution. So this is a no go.) >> > -- A list of frequently asked questions is available at: http://wiki.panotools.org/Hugin_FAQ --- You received this message because you are subscribed to the Google Groups "hugin and other free panoramic software" group. To unsubscribe from this group and stop receiving emails from it, send an email to hugin-ptx+unsubscr...@googlegroups.com. To view this discussion on the web visit https://groups.google.com/d/msgid/hugin-ptx/170638c5-34f5-ca21-b638-e84225acebb0%40suse.cz. For more options, visit https://groups.google.com/d/optout.
