Efficient moving object segmentation algorithm based on the improvement of generalized geodesic active contour model

Abstract

The task of moving object segmentation is to partition an object into non-overlapping regions based on intensity or texture information. However, the conventional segmentation methods suffer from false object segmentation in complex backgrounds and slow convergence. In this paper, we propose an efficient segmentation algorithm for moving object with complicated structures in real video environment. Our novel approach, which integrates an adaptive single Gaussian model (SGM) with a generalized geodesic active contour (GGAC) model, is put forward to detect and segment moving objects in dynamic backgrounds. The proposed algorithm is implemented by level set method to reduce the expensive computational cost of re-initialization of the traditional level set function. By utilizing both spatial and temporal information, this integrated method is robust to complex environments. Experimental results demonstrate desirable segmentation improvement over widely used segmentation algorithms in terms of efficiency and accuracy.