Retinex based on exponent-type total variation scheme (Q1785035)

The authors present a variational model for Retinex, which was developed based on a new image model by decomposing a given image into reflectance, illumination and noises. The model developed in the present paper is coercive, hence it guarantees the existence of the minimizers. The authors also develop an efficient alternating minimization algorithm, where all subproblems can be solved by the closed-form solutions. Several numerical results illustrate the advantages of this model over the existing method for Retinex applications. These numerical experiments include tests of Gaussian noises, stable performance for noises, stable performance for intensity inhomogeneities, extension to color images/impulsive noise/Poisson noise, and convergence analysis.