Vector-Valued Image Interpolation by an Anisotropic Diffusion-Projection PDE

We propose a nonlinear image interpolation method, based on an anisotropic diffusion PDE and designed for the general case of vector-valued images. The interpolation solution is restricted to the subspace of functions that can recover the discrete input image, after an appropriate smoothing and sampling. The proposed nonlinear diffusion flow lies on this subspace and its strength and anisotropy effectively adapt to the local variations and geometry of image structures. The derived model efficiently reconstructs the real image structures, leading to a natural interpolation, with reduced blurring, staircase and ringing artifacts of classic methods. This method also outperforms other existing PDE-based interpolation methods. We present experimental results that prove the potential and efficacy of the method as applied to graylevel and color images.