Non-reference image quality assessment based on deep clustering

作者：

Highlights：

• The proposed framework is adaptive for IQA on images with varying size.

• We incorporate a contract autoencoder to loss function to ensure that the results of IQA has a better stability, robustness, and anti-disturbance.

• Provide that: our proposed CDEC achieves a better image quality evaluation performance than the three ablation models.

摘要

•The proposed framework is adaptive for IQA on images with varying size.•We incorporate a contract autoencoder to loss function to ensure that the results of IQA has a better stability, robustness, and anti-disturbance.•Provide that: our proposed CDEC achieves a better image quality evaluation performance than the three ablation models.

论文关键词：Deep clustering,Quality evaluation,Feature extraction,Contracted autoencoder

论文评审过程：Received 15 June 2019, Revised 14 December 2019, Accepted 3 January 2020, Available online 13 January 2020, Version of Record 27 February 2020.

论文官网地址：https://doi.org/10.1016/j.image.2020.115781