05-27-2021, 10:14 PM
Hello,
I work with photogrammetry, am interested to experiment with using separate processes of RAW, the conventional for projecting color to the model and outputting texture maps, the other for scene reconstruction itself. I'm unsure how any SfM engine will deal with an image that hasn't been demosaic'd (or deBayered), but what I stand to gain in improved resolution of the geometry is worth investigating. Can anyone suggest how this might be made to work? Simply the RAW data saved to 16-bit TIFF holds promise. I could then apply sharpening, though I'm unsure if sharpening won't like the Bayer pattern of 2 green pixels for every one red and one blue pixel. I'm reading about pan-sharpening, which accomplishes exactly that, but is part of a commercial package that includes a host of features I don't need, and is a Python scripted thing over cloud-based that probably costs a ton, as it relates to processing global imagery. Many thanks for ideas.
I work with photogrammetry, am interested to experiment with using separate processes of RAW, the conventional for projecting color to the model and outputting texture maps, the other for scene reconstruction itself. I'm unsure how any SfM engine will deal with an image that hasn't been demosaic'd (or deBayered), but what I stand to gain in improved resolution of the geometry is worth investigating. Can anyone suggest how this might be made to work? Simply the RAW data saved to 16-bit TIFF holds promise. I could then apply sharpening, though I'm unsure if sharpening won't like the Bayer pattern of 2 green pixels for every one red and one blue pixel. I'm reading about pan-sharpening, which accomplishes exactly that, but is part of a commercial package that includes a host of features I don't need, and is a Python scripted thing over cloud-based that probably costs a ton, as it relates to processing global imagery. Many thanks for ideas.
