Object detection and features extraction in video frames using direct thresholding

Abstract

This paper. presents a method for objects detection and features extraction in static video imagery that operates on color/gray-scale frames grabbed by common digital cameras or readily available images from external sources. Segmenting objects is achieved by a technique proposed here and named as direct thresholding (DTh) with background extracted through applying morphological background estimation scheme (MBES). The accommodating hardware is represented by a dedicated P4 PC-based image processing environment, whereas the intrinsic software has been simulated and validated using MATLAB7.3 platform. Human visual perceptual inspection plus histogramming showed an appreciable level of performance through a specific experimentation of multiple-phases. The envisaged diverse fields of application of this method may involve: traffic flow measurement of vehicles and pedestrians; athletic and dancing performance evaluation; public, private, and military security, and so on.