Locally and multiply distorted image quality assessment via multi-stage CNNs

作者：

Highlights：

• We build three image datasets of local distortion, multiple distortion and single distortion.

• Local and multiple distortion classifier will distinguish the locally distorted images from multiply distorted images.

• Quality distortion type classifier will divide the distorted images into three types of image quality distortions.

• Image quality assessment framework based on Multi-stage CNNs performs well on evaluating the quality of both locally and multiply distorted images.

摘要

•We build three image datasets of local distortion, multiple distortion and single distortion.•Local and multiple distortion classifier will distinguish the locally distorted images from multiply distorted images.•Quality distortion type classifier will divide the distorted images into three types of image quality distortions.•Image quality assessment framework based on Multi-stage CNNs performs well on evaluating the quality of both locally and multiply distorted images.

论文关键词：Image quality assessment,Locally distorted image,Multiply distorted image,Convolutional neural network

论文评审过程：Received 29 July 2019, Revised 2 November 2019, Accepted 24 November 2019, Available online 30 November 2019, Version of Record 6 May 2020.

论文官网地址：https://doi.org/10.1016/j.ipm.2019.102175