Defocus-aware Dirichlet particle filter for stable endoscopic video frame recognition

作者：

Highlights：

• We propose a method for smoothing recognition results of NBI endoscopic video frames.

• It is robust to defocused frames by using defocus information extracted from frames.

• We develop a particle filter with the Dirichlet distribution.

• The Rayleigh distribution is used for the defocus information and the likelihood.

• Experimental results are shown with synthetic and real NBI endoscopic videos.

摘要

Highlights•We propose a method for smoothing recognition results of NBI endoscopic video frames.•It is robust to defocused frames by using defocus information extracted from frames.•We develop a particle filter with the Dirichlet distribution.•The Rayleigh distribution is used for the defocus information and the likelihood.•Experimental results are shown with synthetic and real NBI endoscopic videos.

论文关键词：Particle filter,Dirichlet distribution,Defocus information,Endoscopy,Colorectal cancer

论文评审过程：Received 21 April 2015, Revised 17 March 2016, Accepted 17 March 2016, Available online 18 March 2016, Version of Record 13 April 2016.

论文官网地址：https://doi.org/10.1016/j.artmed.2016.03.002