Motion-based object segmentation using hysteresis and bidirectional inter-frame change detection in sequences with moving camera

作者：

Highlights：

• Inter-frame change detection through bidirectional global motion compensated error.

• Automatic optimal weight selection for hysteresis weighted mean thresholding of error images.

• Temporal consistency using adaptive filtering.

• Evaluation using manually created ground truth has shown superior segmentation performance.

• Robustness under various H.264/AVC coding conditions.

摘要

•Inter-frame change detection through bidirectional global motion compensated error.•Automatic optimal weight selection for hysteresis weighted mean thresholding of error images.•Temporal consistency using adaptive filtering.•Evaluation using manually created ground truth has shown superior segmentation performance.•Robustness under various H.264/AVC coding conditions.

论文关键词：Inter-frame change detection,Object segmentation,Hysteresis,Global motion estimation

论文评审过程：Received 10 October 2012, Revised 10 September 2013, Accepted 10 September 2013, Available online 1 October 2013.

论文官网地址：https://doi.org/10.1016/j.image.2013.09.008