Object localization and Tracking Using Background Subtraction and Dual-Tree Complex Wavelet Transform

As seen, the object localization is coupled to many vision applications such as tracking, activity recognition region, security concern, etc. Therefore, segmenting the region of interest to assert the best detection of the target in the sequence of frames is the primary aim of this research. This paper presents an algorithm that detects and tracks the moving object in complex video sequence using the background subtraction and wavelet transform. The work proposes an adaptive background model based on clustering method for regularizing the objection extraction phase. Afterward, it computes the energy of the moving mask using the wavelet coefficient and updates the position of the object by matching this energy to that of moving mask corresponding to next frame. The work also com-pares qualitative and quantitative performance of the proposed method with other existing state-of-the-arts motion detection methods.