Experiment of software binning to improve the SNR of images.
Idea is to effectively reduce the 24MP resolution of D7100 sensor down to 6MP.
Hence, although resolution is reduced each logical pixel gathers more signal.
There is also the fact that higher sensor resolution does not necessarily improve image resolution due to the limitations imposed by the diffraction of atmosphere (see Airy disks, Nyquist etc.). One of the great ironies is that DSLR pixels typically each have two green photosites and only one blue and red - green being the least useful colour for deep-sky images. (As is usaully the case, I'm sure that I'll find out theres a simple built-in way to do this as a batch in PixInsight after I've spent weeks doing it manually
).
Nikon D7100 Celestron C9.25, 111 subs ISO6400 3s.
Using PixInsight...
Original raw NEF files loaded, changed to 32bit floating point samples.
Downsampled by 2 (average mode) to give 2x2 software binning.
I.e. each sample is combination of readout from 4 camera pixels (or 16 photosites). If the sensor supported binning then the values would have been added rather than averaged. However, averaging is not losing information as we are using a much wider sample range then the original.
Binned images saved as FITS format files.
Registered the binned images using middle frame as reference.
Integrated registered images - Average, Additive with scaling, Noise evaluation.
Scale estimage = MAD.
Generated 64bit image.
Cropped off image edges.
Stretched in histogram transformation.
Applied background neutrailisation.
Used automatic background extractor.
Stretched slightly in histogram transformation again.
HDR wavelets: 6 layers, 2 iterations (not inverted), small scale 4, lightness mask, deringing.
ACDNR: with lightness mask to protect stars and nebula.
SCNR to remove last bits of green.
Colour calibration: main nebula area as reference, set limits based on max b/g levels.
Exported to LR as 16bit TIF (max depth LR with load properly).
Final adjustments of curves and levels.
Comments