Here's another attempt at data fusion between JWST and Hubble.
I feel like there is no perfect way to do it, how to represent in visible colors something that is not visible ?
I tried a deep learning framework from Hui Li et al. 2018 : « Infrared and Visible Image Fusion using a Deep Learning Framework ». I found it very conservative when it comes to data. NIRcam, MIRI and HST are decomposed into base parts and detail content. Then the base parts are fused by weighted-averaging. It assembles the details without one dataset overwriting another.
Processing sequence JWST NIRcam
Linear Pattern Substraction, Star alignement, RGB: f360m Maroon, f355m Red, f300m Green, f200w Blue. DBE, Mure Denoise, Deconvolution, TGV, masked stretch, ACDNR, Curves.
Processing sequence JWST MIRI
Linear Pattern Substraction, Dynamic alignment, Superstack of all filters using the Robust Chauvenet Rejection method. DBE, Masked Stretch, local histogram, Curves
Processing sequence of Hubble
Star Alignment, Mure Denoise, RGB as: NIRcam superstarck for maroon, f814w Red, f555w Green, f438w Cyan, f336w Blue. DBE, Deconvolution, ColorCal, TGV, Masked Stretch, ACDNR, Local Hist, Photoshop for cosmetic corrections, filling the gap with noise.
Deep Learning data fusion
Fusion of MIRI and NIRcam superstacks. Fusion of JWST & HST superstacks.
LRGB with the fusion as Luminance and Hubble as RGB.
Comments