A New Video Object Segmentation Algorithm by Fusion of Spatio-temporal Information Based on GMM Learning

In the field of surveillance, Effective and rapid video object segmentation is a key technology for video analysis and processing. For the complex scene and noise that affect segmentation issue in the fixed occasion, on the base of classic Gaussian Mixture Background Model (GMM), a new algorithm named the fusion of Spatio-Temporal based on GMM is proposed for video object segmentation, which classifies for each pixel in Time and Space scales. Firstly, the algorithm constructs dynamically Gaussian Mixture Background Model for each pixel and segment foreground objects through background subtraction. Secondly, the algorithm detects synchronously the neighborhood statistic feature of each pixel through two lemmas. Finally, a result is produced using the spatial segmentation coupling with the temporal segmentation by “and” operator. Experiments show that our proposed algorithm can segment the moving object effectively and quickly from video sequences and has stronger robustness application prospect contrasted with other algorithms.