GBVS360, BMS360, ProSal: Extending existing saliency prediction models from 2D to omnidirectional images
作者：
Highlights：
• Three open source framework BMS360, GBVS360 and Projected Saliency are presented.
• Each model address saliency maps prediction for 360 degree images.
• An adaptive equatorial prior was designed allowing offsets in the pitch’s position.
• The importance of equatorial prior was found dependent of the use case.
• The paper study features interactions while computing saliency map in 360 images.
• A FOV of 150 degree is recommended for direct adaptation of existing models.
摘要
•Three open source framework BMS360, GBVS360 and Projected Saliency are presented.•Each model address saliency maps prediction for 360 degree images.•An adaptive equatorial prior was designed allowing offsets in the pitch’s position.•The importance of equatorial prior was found dependent of the use case.•The paper study features interactions while computing saliency map in 360 images.•A FOV of 150 degree is recommended for direct adaptation of existing models.
论文关键词：Visual attention,Omnidirectional images,Computational model,Feature activation
论文评审过程：Received 21 September 2017, Revised 12 March 2018, Accepted 15 March 2018, Available online 27 March 2018, Version of Record 30 October 2018.
论文官网地址：https://doi.org/10.1016/j.image.2018.03.006
