Image processing system design usually focuses on optimizing the routines needed to solve the problem at hand. It is usually taken for granted that the number of frames to be processed matches the frame rate of the imaging sensor. However, several factors may reduce the effective frame processing rate. In this article, we focus on the effect of memory management and process synchronization on frame acquisition. As more abstraction layers are added to existing image processing tools software coding becomes easier but the control over the image acquisition process decreases. For high-end systems this may not be a problem but for systems with lower processing power the efficient use of memory and CPU is necessary. In this paper, we present and evaluate several image acquisition approaches in the context of a Hyperspectral imaging system that takes as input RAW images to build a HS data cube.

Keywords: Real Time