Random Spray Retinex based colour constancy for perceptual grouped HDR images

High Dynamic Range Imaging is a inevitable step in machine vision for a high dynamically illuminated environment. In cost-wise aspect it is better to convert the high dynamic image to a low dynamic one with preserving the maximum information with the help of Tone Reproduction Algorithm. Another essential feature in colour based object recognition is colour constancy, especially in identifying metal objects. Many previous works carried out to improve the above mentioned methods. In this paper we propose a cost-effective pre-processing method for colour based object recognition in high dynamic range of illumination environment. Tone Reproduction based on Luminance Driven Perceptual grouping is a powerful method to convert the radiance map image to a low dynamic range by preserving maximum information with a human perception. Random spray Retinex is a modified version of the retinex algorithms in which some sort of redundancy is avoided by replacing paths by random sprays, which provides almost similar colour constancy as a human visual system does.