Normalising an image to perception
I am preparing a service that models eye tracking from an image. For curiosity’s sake: an example of normalising a picture to a person’s foveal vision.
The lighter an area, the more strongly it is perceived by vision. The model is trichromacy (full colour vision). The lighting of the areas is logarithmic.
And here is the model based on this normalised representation:
The main areas are shown correctly (according to the tests), but the model is biased toward the left corner: somewhere there is an error in the algorithm. We shall fix it)
UPD
After a series of publications on modelling eye-tracking I have 8 targets with real heat maps from a Tobii reader, on which I test my algorithms. To be honest, I do not have enough of them. If there is an opportunity to carry out research on this question — I shall be glad.