Unsupervised feature learning framework for no-reference image quality assessment

By Peng Ye, J. Kumar, Le Kang, David S. Doermann. 2012

867 citations recorded by the source corpus.

How researchers see this paper

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

1. Kai Zhang
2. Radu Timofte
3. Weixia Zhang
4. Lei Zhang
5. Kede Ma
6. Chen Change Loy
7. Wangmeng Zuo
8. Luc Van Gool

Citations in the corpus

• The Perception-Distortion Tradeoff (2017, background): “[25, 26, 33, 27, 45, 16, 23], which quantify the perceptual quality of an image without a corresponding original image.”
• Learning a No-Reference Quality Metric for Single-Image Super-Resolution (2016, methodology): “In [24, 25], features learned from auxiliary datasets are used to quantify the natural image degradations as alternatives of statistical properties.”
• Blind Image Quality Estimation via Distortion Aggravation (2018, background): “Ten state-of-the-art blind IQA measures act as competitors in this paper, including: (1) NSS based measures, i.e., DIIVINE [15], BLIINDS2 [16], BRISQUE [17], NIQE [28], and ILNIQE [19]; (2) learning based measures, i.e., CORNIA [29] and HOSA [30]; (3) human vision based and some other measures, i.e.”
• KADID-10k: A Large-scale Artificially Distorted IQA Database (2019, unknown): “I QA BIQI [24] 0.460 0.431 0.299 BLIINDS-II [25] 0.559 0.527 0.375 BRISQUE [26] 0.554 0.519 0.368 CORNIA * [27] 0.580 0.541 0.384 DIIVINE [28] 0.532 0.489 0.341 HOSA * [29] 0.653 0.609 0.438 SSEQ [30] 0.463 0.424 0.295 InceptionResNetV2 (fine-tune) 0.734 0.731 0.546 * To reduce time complexity,…”
• Learning without Human Scores for Blind Image Quality Assessment (2013, methodology): “D histgram CORNIA [27] 3.246 O(Nd2K) d: blocksize, K: codebook size QAC 0.189 O(N(h2/s2)) s: block step; h: filter window size BRISQUE [10] 0.176 O(Nd2) d: filter window size BIQI [13] 0.076 O(N) judge the quality of an input image online.”
• NTIRE 2017 Challenge on Single Image Super-Resolution: Dataset and Study (2017, methodology): “CORNIA is a perceptual measure which does not use a reference image.”
• RankIQA: Learning from Rankings for No-Reference Image Quality Assessment (2017, methodology): “The codebook method [40, 39] combines different features instead of using local features directly.”
• Quality Assessment of In-the-Wild Videos (2019, methodology): “ates the video frame length and the resolution. Method 240frs@540p 364frs@480p 467frs@720p 450frs@1080p BRISQUE [27] 12.6931 12.3405 41.2220 79.8119 NIQE [29] 45.6477 41.9705 155.9052 351.8327 CORNIA [51] 225.2185 325.5718 494.2449 616.4856 VIIDEO [28] 137.0538 128.0868 465.2284 1024.5400 VBLIINDS [”
• From Patches to Pictures (PaQ-2-PiQ): Mapping the Perceptual Space of Picture Quality (2019, background): “Examples include BRISQUE [10], NIQE [11], CORNIA [13], FRIQUEE [12], which use “handcrafted” statistical features to drive shallow learners (SVM, etc.).”
• Deep Image Deblurring: A Survey (2022, methodology): “…deblurring methods on real-world images, several no-reference metrics have been used, such as BIQI Moorthy and Bovik (2010), BLINDS2 Saad et al. (2012), BRISQUE Mittal et al. (2012a), CORNIA Ye et al. (2012), DIIVINE Moorthy and Bovik (2011), NIQE Mittal et al. (2012b), and SSEQ Liu et al. (2014).”
• Blindly Assess Image Quality in the Wild Guided by a Self-Adaptive Hyper Network (2020, methodology): “In [14], a simple CNN with pooling strategy inherited from [43] is used for quality prediction.”
• Perceptual Quality Assessment of Smartphone Photography (2020, methodology): “We compare the proposed methods with seven existing BIQA models, including BRISQUE [22], DIIVINE [24], CORNIA [39], QAC [38], ILNIQE [40], FRIQUEE [6], and DB-CNN [41].”
• Blind Image Quality Assessment via Deep Learning (2015, unknown): “Mittal et al. [26] presented BRISQUE to predict image quality in the spatial domain, Ye et al. [12] proposed CORNIA to extract features using K-means clustering and applied SVM to map the feature to quality.”
• dipIQ: Blind Image Quality Assessment by Learning-to-Rank Discriminable Image Pairs (2017, methodology): “We re-train BRISQUE [8], DIIVINE [10], and CORNIA [9] on the LIVE database, whose learning tools and parameter settings”
• MUSIQ: Multi-scale Image Quality Transformer (2021, background): “In the past years, researchers proposed many IQA methods: early natural scene statistics based [13, 25, 27, 45], codebook-based [38, 40] and CNN-based [11, 33, 34, 41, 47].”
• Exploring CLIP for Assessing the Look and Feel of Images (2022, background): “Unlike existing NR-IQA methods ( e.g. , MUSIQ) that train different models for different domains ( e.g. , technical quality, aesthet-3 ics quality), CLIP-IQA + needs to store only two prompts for each domain, significantly reducing the storage cost.”
• Group MAD Competition? A New Methodology to Compare Objective Image Quality Models (2016, methodology): “The four distortion types are common in existing IQA databases [27, 23] and many IQA models are claimed to be able to properly handle these distortions [19, 20, 25, 17, 40, 18, 38, 37, 26, 42, 36, 7, 35].”
• A Comparative Study for Single Image Blind Deblurring (2016, background): “While numerous full-reference [36] [37] and no-reference image quality metrics [19] [23] [24] [26] [27] [33] [47] have been proposed, it is unclear whether these metrics can be applied to measure the quality of deblurred images.”