Wavelet-enhanced detection of small/slow object movement in complex scenes

Detecting and tracking moving objects in complex real world scenes is a fundamental process in many applications, including intelligent surveillance, advanced robotics, and human-computer interaction. Moving objects detection means detection and isolation of active objects that take part in the surveyed scene. Yet, the traditional motion detection methods are application-dependent and have limited capabilities in detecting small and slowly moving objects, especially in the presence of illumination variations. This paper aims at figuring out an efficient way to utilize the wavelet analysis in order to enhance the performance of memory-based frame differencing for the detection of small / low contrast moving objects in scenes with different types of illumination variations. Three different ways of utilizing 2-D Discrete Wavelet Transform (DWT) are examined. The proposed method realized in three different variations was evaluated via synthetic motion detection benchmark dataset and realistic scenarios' experiments. Those experiments were performed to assess both the accuracy and performance of the proposed techniques in comparison to the other traditional methods. The comparisons revealed that the proposed method overcome the limitations experienced with the traditional methods, especially in the case of small/low contrast objects moving in environments with varying illumination.