Acceleration of moving object detection in bio-inspired computer vision

Abstract

Computer vision is a great interest field offering relevant information in a wide variety of areas. Different video processing techniques, for instance, allow us to detect moving objects from image sequences of fixed surveillance cameras. Lateral Interaction in Accumulative Computation is a classical bio-inspired method that is usually applied for detecting moving objects in video processing. This method achieves high precision but also requires a high processing time. This paper introduces a parallel code capable of keeping a high performance in terms of accuracy and runtime for the method. For some of the image sequences tested, a speed-up of 67× over the sequential counterpart is achieved.