The problem of inpainting involves reconstructing the missing areas of an image. Inpainting has many applications, such as reconstructing old damaged photographs or removing obfuscations from images. In this paper we present the directional diffusion algorithm for inpainting. Typical diffusion algorithms are bad at propagating edges from the image into the unknown masked regions. The directional diffusion algorithm improves on the regular diffusion algorithm by reconstructing edges more accurately. It scores better than regular diffusion when reconstructing images that are obfuscated by a text mask.