This work deals with lightness and brightness computations and, in particular, the use of retinex-like algorithms for these computations. The retinex algorithm, proposed originally by E.Land, was essentially a one-dimensional computation. Previous extensions to two-dimensions had severe shortcomings, and owing to lack of an efficient solution they were not tested on large-scale images. In this work we produced efficient two-dimensional retinex-like computations, and studied their application to test images. Theoretical and empirical results were then used to analyze the limitations of retinex-like computations.
Back to home page