Illuminant estimation using a particle swarm optimized fuzzy neural network

This paper presents a novel approach for estimating unknown illuminant, i.e., color temperature, of a digitally captured image, depending only on the contents of the image. The novelty of the approach lies in establishment of a fuzzy neural network with its parameters optimized based on appropriately designed training sets using particle swarm algorithm. The estimated temperature may be further utilized to perform color balance for the image. The results indicate strong improvement of the proposed approach over a previous work. A comparative study also suggests that the proposed method is more adaptable to generic images and varieties of illuminants and scenes than other existing methods do.