I have several images (frames from a camera approx. 25,000 images) for which I have one unique lux value ranging from 1 to 110,000). Not all the lux values are covered from 1 to 110,000. Now, I need to predict the lux values in real-time for the same camera once every second, especially predicting the low lux values (1 to 6) are of prime importance.
How do I do it?
I have found one link where there's piecewise relation shown between lighting step (don't understand what the author means by that) and mean lux values here in the second graph (https://docs.microsoft.com/en-us/windows/desktop/SensorsAPI/understanding-and-interpreting-lux-values).
Should I do plain linear regression on these sparse values? If yes, how does that work? Any guide or link will be helpful. Or string together a piecewise linear set of equations to predict the lux values from mean intensities of my frames?