Junjie Wu

Junjie Wu

Auto-generated profile, based on publicly available information

Junjie Wu

School of Computer Science and Technology, Soochow University, Suzhou, China

Research Focus & Metrics

Academic Career Length

9Years (2018-Present)

(Calculated once Scilit publication data is synced: current year − year of first publication + 1)

Representative Publications

View All Publications

10 December 2019 in IEEE Access
Verified
Journal Article
Closed access
CZCongxuan Zhang
Congxuan Zhang
CHChao He
Chao He
ZCZhen Chen
Zhen Chen
WLWen Liu
Wen Liu
MLMing Li
Ming Li
JWJunjie Wu
Junjie Wu

1 October 2018
Verified
Conference Paper
Closed access
JWJunjie Wu
Junjie Wu
XWXuebin Wang
Xuebin Wang
ZCZhen Chen
Zhen Chen
CZCongxuan Zhang
Congxuan Zhang
In order to solve the excessive smoothness caused by existing optical flow algorithms within motion edge regions of images under difficult scenes, such as noise, illumination changes and shadows, occlusion, large displacement, and non-rigid motion, a robust variational optical flow estimation model based on the rolling guidance filter has been proposed in this paper. Firstly, the rolling guidance filter strategy is presented, and the energy function of the rolling guidance filter is designed. Secondly, a non-local total variation with L1 norm (TV-Ll) optical flow computational model based on the rolling guidance filter is constructed. Finally, the energy function is converted into a linear minimization of the TV-Ll optical flow through the multi-resolution pyramid refinement, and the flow field is computed at each layer. The rolling guidance filter is used to optimize the optical flow estimation alternately. The MPI-Sintel and KITTI test sequences are employed to evaluate the proposed algorithm and other state-of-the-art methods, including total variation regularization of local-global optical flow (CLG- TV), classic model with non-local constraint (Classic+ NL), and nearest neighbor fields (NNF). The experimental results showed that the proposed algorithm, compared with other contrast methods, has a better edge protection effect in difficult scenes and motion forms, which effectively improves the accuracy and robustness of the optical flow estimation.