陈志波  (教授)

博士生导师

电子邮箱:

学位:博士

毕业院校:Tsinghua University

   

Predictive Auto-Encoding Network for Blind Stereoscopic Image Quality Assessment

点击次数:

Introduction

Stereoscopic image quality assessment (SIQA) has encountered non-trivial challenges due to the fast proliferation of 3D contents. In the past years, deep learning oriented SIQA methods have emerged and achieved spectacular performance compared to conventional algorithms which are only relied on extracting hand-crafted features. However, most existing deep SIQA evaluators are not specifically built for stereoscopic content and consider little prior domain knowledge of the 3D human visual system (HVS) in network design. In this paper, to better simulate the binocular rivalry phenomenon, we propose a Predictive Auto-encoDing Network (PAD-Net) for blind/No-Reference stereoscopic image quality assessment (NR-SIQA). The proposed encoder-decoder architecture is inspired by the predictive coding theory that the cognition system tries to match bottom-up visual signal with top-down predictions. Besides, we design the Siamese framework to mimic the binocular rivalry in the 3D HVS based on the likelihood and prior maps generated from the predictive coding process. Extensive experimental results on three publicly available stereoscopic image quality databases demonstrate that the devised approach outperforms state-of-the-art algorithms for predicting the perceptual quality of both symmetrically and asymmetrically distorted stereoscopic images with various distortion types.

Paper

Xu J, Zhou W, Chen Z, et al. Predictive Auto-Encoding Network for Blind Stereoscopic Image Quality Assessment[J]. arXiv preprint arXiv:1909.01738, 2019.

Download

To get the source code, Please fill THIS FORM and the source code will be sent to you.


上一条: No-Reference Light Field Image Quality Assessment

下一条: Progressive Image Inpainting with Full-Resolution Residual Network