Mathematical Approaches to Digital Image Inpainting

Image inpainting process is used to restore the damaged image or missing parts of an image. This technique is used in some applications, such as removal of text in images and photo restoration. There are different types of methods used in image inpainting, such as non-inear partial differential equations(PDEs), wavelet transformation and framelet transformation. We studied the usage of the current image inpainting methods and solved the Poisson equation using a five-point stencil method. We used a modified five-point stencil method to solve the same equation. It gave better results than the standard five-point stencil method. Using modified five-point stencil method results as the initial condition, we solved the iterative linear and non-linear diffusion PDE. We considered different types of diffusion conductivity and compared their results. When compared with PSNR values, the iterative linear diffusion PDE method gave the best results where as constant diffusion conductivity PDE gave the worst result. Furthermore, inverse diffusion conductivity PDE had given better results than that of the constant diffusion PDE. However, it was worse than the Gaussian and Lorentz diffusion conductivity PDE. Gaussian and Lorentz diffusion conductivity iterative linear PDE had given a better result for image inpainting. When we use any inpainting technique, we cannot restore the original image. We studied the relationship between the error of the image inpainting and the inpainted domain. Error is proportional to the value of the Greens function. There are two types of methods to find the Greens function. The first method is solving a Poisson equation for a different shape of domain, such as a circle, ellipse, triangle and rectangle. If the inpainting domain has a different shape, then it is difficult to find the error. We used the conformal mapping method to find the error. We also developed a formula for transformation from any polygon to the unit circle. Moreover, we applied the Schwarz Christoffel transformation to transform from the upper half plane to any polygon