A Local-Adaptive Threshold System with DFT Compound Motion Analysis

This paper is concerned with investigating, experiencing, and validating a local adaptive threshold system with compound motion analysis. The motivation here is to introduce a method for analyzing moving objects in outdoor/indoor video frames with respect to: movement detection, objects segmentation, features extraction besides DFT-based velocity computation. The underlying methodology believes in cross-fertilizing the two independent segmentation approaches of background subtraction and temporal frames differencing through a single correlation exhibiting the behavioral-mathematical model of the examined image sequences. This has been interpreted through identifying the image as time-varying functions applicable for processing through 2D Discrete Fourier Transform (DFT). The justification of this method has been revealed through output data, human visual perceptual inspection, plus histogram Ming, showing appreciable accuracy, lower level of noise, and shorter segmentation time in comparison with some available standard techniques. The horizon of applications of the presented method may involve: security control, industry and traffic control, athletic and dancing performance evaluation, surveillance, and general civil and military fields.