DeepSim: Deep similarity for image quality assessment

By Fei Gao, Yi Wang, Panpeng Li, Min Tan, Jun Yu, Yani Zhu. 2017

156 citations recorded by the source corpus.

Abstract This paper studies one interesting problem: how does the deep neural network (DNN) architecture affect the image quality assessment (IQA) performance? In order to find the answer, we propose a novel full-reference IQA framework, codenamed deep similarity (DeepSim). In DeepSim, we first measure the local similarities between the features (produced by a DNN model) of the test image and those of the reference image; Afterwards, the local quality indices are gradually pooled together to estimate the overall quality score. In addition, various factors that may affect the IQA performance are investigated. Thorough experiments conducted on standard databases show that: (1) DeepSim can accurately predict human perceived image quality and outperforms previous state-of-the-art; (2) mid-level representations are most effective for quality prediction; and (3) preprocessing, the restricted linear units and max-pooling operations are beneficial for the IQA performance.

Subfields: Computer Vision and Pattern Recognition, Media Technology, Evaluation & Benchmarks

How researchers see this paper

The researchers whose trusted networks rank it highest, strongest advocate first. Every researcher gets a different order.

In Computer Vision and Pattern Recognition

1. Luc Van Gool
2. Andrea Vedaldi
3. Ming-Hsuan Yang
4. Trevor Darrell
5. Alexei A. Efros
6. Dacheng Tao
7. Jitendra Malik
8. K. Müller

In Media Technology

1. Richard Zhang
2. Eli Shechtman
3. Alexei A. Efros
4. Phillip Isola
5. Jun-Yan Zhu
6. Oliver Wang
7. Daniel Cohen-Or
8. Antonio Torralba

In Evaluation & Benchmarks

1. Richard Zhang
2. Eli Shechtman
3. Phillip Isola
4. Alexei A. Efros
5. Oliver Wang
6. Jun-Yan Zhu
7. K. Müller
8. Wojciech Samek

Citations in the corpus

• The Unreasonable Effectiveness of Deep Features as a Perceptual Metric (2018, background): “to predict visual similarity by training on low-level differences. Concurrent work by Talebi and Milanfar [50] train a deep network in the context of no-reference IQA for image aesthetics. Gao et al. [15] and Amirshahi et al. [3] propose techniques involving leveraging internal activations of deep n”
• PieAPP: Perceptual Image-Error Assessment Through Pairwise Preference (2018, methodology): “, [17, 24,28]), in this paper we only compare against the three that performed the best on established datasets based on their published results [7, 25].”
• Deep Neural Networks for No-Reference and Full-Reference Image Quality Assessment (2016, methodology): “DeepSim [14] extracts”