Monochrome DNG processing on GPU
It's well-known that there are some limitations that we run into when we are working with cameras which have color filter array (CFA). Such cameras have so called "bayer pattern" - set of color filters on image sensor pixels. As soon as we need color image, we have to interpolate missing values of color components and this is the reason to get some errors on final image. This is how we loose color precision and sharpness after interpolation. Quantum efficiency of each pixel is also a problem, because CFA cut-off filter removes significant part of visual spectrum and finally we get slower signal from each pixel.
We can avoid these problems if we utilize monochome cameras. These cameras won't give us precise colors, but they can offer better signal to noise ratio, better sharpness and we need to do less computations to get final image from raw. If we don't bother about colors in any particular task, it means that we can try to work with monochrome cameras.
We've implemented GPU-based image processing pipeline for high performance software to handle monochrome DNG footages at Fast CinemaDNG Processor software. As soon as monochrome data has less size and we don't need demosaicing, color correction and color grading, total performance is significantly better in comparison with color DNG with the same bit depth and resolution.