Another look at Land´s Retinex algorithm

D. H. Brainard and B. A. Wandell (1986) have criticized the Retinex algorithm as being flawed in its adverse sensitivity to the composition of the image´s scene. The author considers whether and to what extent the Retinex algorithm can provide useful information for automated processes that rely on color identification in spite of changing conditions in the scene. The performance of the algorithm was tested using simulated images. The objective was to understand how well it performed through various changes in the makeup and properties of the image. The results, in general, show that the algorithm is tolerant of changes in the makeup of the image