I wonder if there would be any benefit for ML models to non linearly distorting image pixel statistics into a true Gaussian distribution instead of just the traditional scale and bias.
Non-linear image distortion for Gaussian pixel statistics in ML models
By
–