Human, animal or object surface`s color image correcting method, involves forcing output of neuron of median layer in neural network, during propagation steps of learning phase, to luminance value of output pixel

摘要

The method involves using pairs of learning pixels subjected to illumination changes when a learning pixel of a pair is applied to an input of a neural network (RES). An output of a neuron (N1) of a median layer in the neural network has its output forced, during propagation steps of a learning phase, to a luminance value (L) of an output pixel. The neuron has its output forced to a predetermined luminance value after the learning phase, while the pixels of the color image to be corrected are applied to the input of the network to obtain corrected pixels at the output. Independent claims are also included for the following: (1) a method for obtaining a color invariant for a given color (2) a color image correcting device implementing a color image correcting method (3) a computer program comprising instructions for implementing a color image correcting method.