Implementation of a wide-angle lens distortion correction algorithm on the cell broadband engine

Abstract

Wide-angle lenses are often used in scientific or virtual reality applications to enlarge the field of view of a conventional camera. Wide-angle lens distortion correction is an image warping application which transforms the distorted images back to the natural-looking, central perspective space. This application is characterized by a non-linear streaming memory access pattern that makes main memory bandwidth a key performance limiter. This paper presents the implementation, optimization and evaluation of a wide-angle lens distortion correction algorithm on the Cell Broadband Engine. Cell is a heterogeneous multi-core processor that has been architected to accelerate applications with large degree of thread- and data-level parallelism. We alleviate the 'memory wall' inefficiencies by applying source level optimizations such as tiling to better utilize the on-chip memory of the SPE, and maximize data reuse within a frame of pixel data. Using these transformations on the Cell processor, we are able to achieve a 7.27x speedup compared to a Core2 Duo processor, and enable potential applications such as real-time correction for video teleconferencing using cameras with wide-angle lenses. To the best of our knowledge, this is the first paper to describe the mapping and optimization of an image warping application to the Cell BE architecture. Copyright 2009 ACM.