Nonlinear diffusion at your fingertips: Theory and mobile applications

Abstract

Through affordable smartphones and lower costs for mobile internet access, mobile communication by moving and still pictures has become a commonplace. However, while most phones offer a camera for taking pictures, the image processing capabilities are relatively sparse. The challenges that arise with real-time image handling are not yet countered by complex techniques. Because of the limited computation power, simple algorithms were used. In this paper, we will show that the technique of nonlinear diffusion can be used in (near) real-time on most smartphone's GPUs if being implemented in an OpenGL ES shader. The presented implementation is then used for several image-related tasks, namely denoising, deblocking and upscaling directly on the phone. As the performance evaluation shows, anisotropic diffusion offers a nice framework for complex mobile image handling in several arising problem areas.