A Novel Video-based Human Object Extraction Method
Download as PDF
Qiuyan Li, Yan Ma, Hui Huang, and Yuping Zhang
It is one of the important tasks for human data acquisition to extract the human object with the specific action. However, the accuracy of object extraction will be influenced by the shadow in an image as well as the non-standard human action. To address this issue, we propose a novel human object extraction method. First, we extract the moving object contour with the combination of three-frame-difference method and mixture gaussian model. Next, we remove the shadows from the image with multi-feature fusion method. Then, we extract the human skeleton with distance transformation. Finally, we select the image according to the angle of skeleton. The experimental results show that the proposed method can accurately extract the human object with the specific action from the image.
Human data acquisition, Object extraction, mixture gaussian model, Distance transformantion, Three-frame-difference method