A primary-secondary foreground segmentation method with window series PCA de-noising

Noise characteristic and motion properties of different foreground objects under various weather conditions are analyzed for outdoor videos, and a primary-secondary foreground segmentation method is proposed. A window series PCA algorithm, combined with the Gaussian mixture model, is used to model the videos after de-nosing and segmenting all foreground objects primarily. After that, the probabilities of the overlapped regions are calculated to describe the motion properties of different objects, and a second segmentation step is carried out to extract the interesting objects. Finally, the uninteresting objects, such as raindrops and snowflakes, are treated via a background-inpainting step to improve the video quality. Experimental results show that our proposed method can effectively reduce noise, diminish the interference of rain or snow, and enhance video effects.