Abstract
In this paper, we propose a new functional model for denoising and enhancing the chromaticity component of color images. This model separates the color data into chromaticity and brightness. The three color channels of each pixel are regarded as a 3-dimensional vector, and the unit directional vector and its magnitude represent the chromaticity and the pixel brightness separately. The brightness component is processed by an improved forth-order PDE model which was proposed by authors in [8]. For the chromaticity component, it is considered as an embedded surface on Riemannian manifold. A physical quantity in the form of vector product and the unit norm restriction on chromaticity component are introduced into the energy functional of the proposed model. Furthermore, we give a detailed derivation of Euler-Lagrange equation for the energy functional. Numerical experimental results show that the model can preserve the chromatic characteristics while removing Gaussian noise and salt-pepper noise.