Automatic white balancing by hybrid neural model

摘要

This paper presents an automatic white balancing (AWB) study achieved by a hybrid neural model which includes self-organizing map (SOM) neural network, feed-forward neural network (FFNN), and von Kries chromatic adaptation. The illuminated effect of an image is expected to be greatly eliminated and improved. In the hybrid neural model proposed, SOM neural network partitions the nonlinear color space into a uniform topology space which can be used for illuminant estimation by FFNN neural network. To demonstrate the efficiency of the model proposed, several experiments were executed. From the experimental results shown, the effect of illuminant has been improved effectively as compared with the works done before.