Mapping colorbar of one image to another to obtain physical values.

I have an array of values with numbers and NaN of physical values where detectors are located, and where they aren't, respectively.

I used imagesc and colorbar to plot the data as an image map.

I have another image having a full map of values where the NaN have been replaced by colours (representing predicted values).

How can I obtain the numerical values of the predicted values (from RGB to my physical variable), from this complete image, relative to the magnitude of the original image with NaN's and some known values.

Note that, the complete map can have values more than any value in the original image with NaNs...

[Ibrahim Patel]

As a visualisation of what I am trying to ask, please see my comment here .

[Ibrahim Patel]

Ok, so I tried reading index values of both images and used a proportional relation to calculate the physical parameter magnitude of the predicted values. There is some error on account of using rgb2ind with drifting, but it is in a tolerable range.

Now, I'd like to know how to automatically get index values for each position, given x and y position values, instead of manually seeking the index of each colored region.