Multi-layer fusion network for blind stereoscopic 3D visual quality prediction

作者：

Highlights：

• Training patches with more saliency information can improve the accuracy of prediction results.

• We fuse low-, middle-, and high-level features to predict the visual quality of S3D images.

• The shared trunk network predicts weights and qualities of S3D image patches.

摘要

•Training patches with more saliency information can improve the accuracy of prediction results.•We fuse low-, middle-, and high-level features to predict the visual quality of S3D images.•The shared trunk network predicts weights and qualities of S3D image patches.

论文关键词：Stereoscopic 3D image,Visual quality prediction,Dual-stream,Fusion network,Binocular vision

论文评审过程：Received 29 May 2020, Revised 29 October 2020, Accepted 30 November 2020, Available online 2 December 2020, Version of Record 4 December 2020.

论文官网地址：https://doi.org/10.1016/j.image.2020.116095