Group MAD Competition? A New Methodology to Compare Objective Image Quality Models

By Kede Ma, Q. Wu, Zhou Wang, Zhengfang Duanmu, Hongwei Yong, Hongliang Li, Lei Zhang. 2016

106 citations recorded by the source corpus.

Subfields: Computer Vision and Pattern Recognition, Media Technology

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. Dacheng Tao
2. Wangmeng Zuo
3. Luc Van Gool
4. Kede Ma
5. Lei Zhang
6. Kai Zhang
7. Radu Timofte
8. Ming-Hsuan Yang

In Media Technology

1. Kede Ma
2. Wangmeng Zuo
3. Kai Zhang
4. Dacheng Tao
5. Lei Zhang
6. Weixia Zhang
7. Radu Timofte
8. Wentao Liu

Citations in the corpus

• Pixel Recursive Super Resolution (2017, neutral): “Many methods exist for quantifying image similarity that attempt to measure human perception judgements of similarity [45, 46, 28].”
• TAC-GAN - Text Conditioned Auxiliary Classifier Generative Adversarial Network (2017, methodology): “To show that our model produces very diverse sample, we used the MS-SSIM [13, 25].”
• End-to-End Blind Image Quality Assessment Using Deep Neural Networks (2018, methodology): “gy [2] on the Waterloo Exploration Database [29].”
• KonIQ-10k: An Ecologically Valid Database for Deep Learning of Blind Image Quality Assessment (2019, methodology): “Ma et al. [24] created a database with 4,744 pristine images and 94,880 distorted images to validate their proposed mechanism called group MAximum Differentiation (gMAD) competition.”
• Blind Image Quality Assessment Using a Deep Bilinear Convolutional Neural Network (2019, methodology): “Differentiation (gMAD) competition [25], we observe that DB-CNN is more robust than the most recent CNN-based”
• Blindly Assess Image Quality in the Wild Guided by a Self-Adaptive Hyper Network (2020, methodology): “Although these IQA methods have achieved great performance improvement on synthetic databases, challenge exists when facing large scale data [25, 28], indicating the problem of content variation still not well managed.”
• dipIQ: Blind Image Quality Assessment by Learning-to-Rank Discriminable Image Pairs (2017, methodology): “2) Base Feature: We adopt CORNIA features [9] to represent test images because they appear to be highly competitive in a recent gMAD competition on the Waterloo Exploration Database [19].”
• PieAPP: Perceptual Image-Error Assessment Through Pairwise Preference (2018, methodology): “Reference images: The proposed dataset contains 200 unique reference images (160 reference images are used for training and 40 for testing), which are selected from the Waterloo Exploration Database [36,37] because of its highquality images.”
• KADID-10k: A Large-scale Artificially Distorted IQA Database (2019, unknown): “of Ratings Subjective study Database Year Content distorted images Distortion type distortion types rated images per image environment IVC [5] 2005 10 185 artificial 4 185 15 lab LIVE [6] 2006 29 779 artificial 5 779 23 lab CSIQ [7] 2009 30 866 artificial 6 866 5∼7 lab TID2013 [2] 2013 25 3,000 arti”