摘要

This paper presents a new tracking method based on local matting approach for accurate and robust night fusion image tracking. Firstly, frame difference is applied to produce the coarse target, which helps to generate foreground and local background models. These models provide sufficient and accurate restriction for matting, which makes matting applicable in tracking process. Then accurate boundaries of the target can be extracted from matting results. Finally tracking results is used to construct and update foreground and background models in time. Extensive experiments show that our scheme can largely avoid model drift and show good performance when targets are deforming and quick moving.