Video saliency detection incorporating temporal information in compressed domain

作者：

Highlights：

• We propose a compressed video saliency model for which few attention is given to.

• The characteristics of codec are considered to remove the effects of QP.

• We use K-means clustering to statistically distinguish the motion attention level.

• The visual window is built to strengthen the contrast of features.

• The variance-like fusion method is used to compute the video saliency map.

摘要

Highlights•We propose a compressed video saliency model for which few attention is given to.•The characteristics of codec are considered to remove the effects of QP.•We use K-means clustering to statistically distinguish the motion attention level.•The visual window is built to strengthen the contrast of features.•The variance-like fusion method is used to compute the video saliency map.

论文关键词：Compressed domain,Video saliency detection,Visual window,Motion importance factor

论文评审过程：Available online 12 August 2015, Version of Record 18 November 2015.

论文官网地址：https://doi.org/10.1016/j.image.2015.07.013