Consumer graphics cards for fast image processing based on the Pixel Shader 3.0 standard

Optical inspection and metrology for industrial applications often require fast image processing. Today this is mostly reached by special developed hardware, which represents a relevant cost factor in image processing systems. Furthermore, such special solutions often cannot be easily adapted to new requirements. Otherwise, CPU based systems feature a good programmability but often not enough processing power for real time purposes. The required computational performance and flexibility can be provided by the latest issue of consumer graphics cards for personal computers equipped with a fourth generation Graphics Processing Unit (GPU) supporting the Pixel Shader 3.0 standard. The GPU contains freely programmable units for geometrical and mathematical processing of textured objects by shader programs. The objective of this paper is to show that the 3.0 standard shaders are now ready to be integrated in industrial environments for real time imaging applications. To this end, we focus on common image processing algorithms, such as linear and nonlinear filters. A benchmark has been implemented to compare the speed of GPU shader algorithms to those running on CPU systems.