Research on Video Object Segmentation Algorithm Based on Background Reconstruction
Fengmin Cheng · 2015
Aiming at the problem of occlusion and shadow in the existing video object segmentation algorithm, a new video object segmentation based on background reconstruction is proposed.In the time domain, the motion area of a single frame image is obtained, and the method of changing the template with the frame difference or the two frame difference map is discarded.By using the two frame difference to obtain the motion information of the single frame, the threshold is obtained by the frame difference mask image, thus improving the accuracy of background reconstruction.The threshold value of the process is obtained by LHS method.The method can generate adaptive threshold in time.Finally, a method of eliminating the shadow is proposed.