Evaluation and FPGA Implementation of Sparse Linear Solvers for Video Processing Applications

In this work, we address sparse linear solvers for real-time video applications.

Sparse linear systems are commonly used in video processing applications, such as edge-aware filtering or video retargeting. Due to the 2D nature of images, the involved problem sizes are large and thus solving such systems is computationally challenging. In this work, we address sparse linear solvers for real-time video applications. We investigate several solver techniques, discuss hardware trade-offs, and provide FPGA architectures and implementation results of a Cholesky direct solver and of an iterative BiCGSTAB solver. The FPGA implementations solve 32K32K matrices at up to 50 fps and outperform software implementations by at least one order of magnitude.