Adaptive Image Colorization with Total Variation and Partial Differential Equation

An adaptive image colorization algorithm based on total variation and partial differential equations (PDE) is proposed to overcome problems of color blurring near edges in some colorization algorithms. A partial differential equation is established through minimization of a functional. An adaptive method is then used to select the model parameters. By choosing suitable diffusion coefficients related to the image properties, we can improve efficiency of diffusion in smooth image regions, and continuity of colors near edges. A colorized image is obtained by numerically solving the PDE with a finite-difference approach. Compared with some other PDE methods, the proposed method can produce colorized images with better visual quality.