RankIQA: Learning from Rankings for No-Reference Image Quality Assessment

By Xialei Liu, Joost van de Weijer, Andrew D. Bagdanov. IEEE International Conference on Computer Vision, 2017

516 citations recorded by the source corpus.

We propose a no-reference image quality assessment (NR-IQA) approach that learns from rankings (RankIQA). To address the problem of limited IQA dataset size, we train a Siamese Network to rank images in terms of image quality by using synthetically generated distortions for which relative image quality is known. These ranked image sets can be automatically generated without laborious human labeling. We then use fine-tuning to transfer the knowledge represented in the trained Siamese Network to a traditional CNN that estimates absolute image quality from single images. We demonstrate how our approach can be made significantly more efficient than traditional Siamese Networks by forward propagating a batch of images through a single network and backpropagating gradients derived from all pairs of images in the batch. Experiments on the TID2013 benchmark show that we improve the state-of-theart by over 5%. Furthermore, on the LIVE benchmark we show that our approach is superior to existing NR-IQA techniques and that we even outperform the state-of-the-art in full-reference IQA (FR-IQA) methods without having to resort to high-quality reference images to infer IQA.

Subfields: Computer Vision and Pattern Recognition, Media Technology

View Paper: https://www.semanticscholar.org/paper/50fa828ea098abc91bb6e5750657a7d4cf85e07d

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

In Media Technology

Citations in the corpus

TOPIQ: A Top-Down Approach From Semantics to Distortions for Image Quality Assessment (2023, background): “[42] introduced a ranking loss for pretraining networks with synthetic data.”
GIQA: Generated Image Quality Assessment (2020, methodology): “763 RankIQA[40] 0.”
Learning Metrics From Teachers: Compact Networks for Image Embedding (2019, background): “However, both hard and semi-hard negative mining have a high computational cost, which led several authors to limit the hard negative mining process to the current mini-batch [18, 31, 32, 36].”
Leveraging Unlabeled Data for Crowd Counting by Learning to Rank (2018, methodology): “However, in [20] the authors show that it is computationally advantageous (and sometimes leads to better minima) to send the images in a batch through a single branch and combine them when computing the ranking loss.”
Domain Fingerprints for No-Reference Image Quality Assessment (2019, methodology): “For pre-training, as in the previous studies [4], [44], we reproduce 17 out of a total of 24 distortion types in TID2013 and apply them to the 840 high-quality images.”
From Patches to Pictures (PaQ-2-PiQ): Mapping the Perceptual Space of Picture Quality (2019, background): “Several deep NR models [42, 43, 44, 45, 46] have also been created that yield state-of-the-art performance on legacy synthetic distortion databases [3, 4, 26, 41], by pretraining deep nets [47, 48, 49] on ImageNet [50], then fine tuning, or by training on proxy labels generated by an FR model [45].”
Metric Learning for Novelty and Anomaly Detection (2018, methodology): “In practice, we have not implemented a two-branches Siamese network but followed recent works [18, 28] which devise a more efficient approach to minimize losses traditionally computed with Siamese networks.”
No-Reference Image Quality Assessment via Transformers, Relative Ranking, and Self-Consistency (2021, background): “These approaches [61, 62, 38, 39] address NR-IQA as a learning-to-rank problem, where the relative ranking information is used during the training.”
KonIQ-10k: An Ecologically Valid Database for Deep Learning of Blind Image Quality Assessment (2019, methodology): “Liu et al. proposed the RankIQA method [40].”
Blind Image Quality Assessment via Vision-Language Correspondence: A Multitask Learning Perspective (2023, background): “, Gaussian noise and JPEG compression) are of primary concern, patchwise training [4], quality-aware pre-training [32, 37, 76], and learning from noisy pseudolabels [2, 38, 67] are practical training tricks with less (or no) reliance on MOSs.”
Blindly Assess Image Quality in the Wild Guided by a Self-Adaptive Hyper Network (2020, background): “Noticing the limited size of training data from existing IQA databases, [24] and [26] proposed to generate vast training samples by labeling their quality rank instead of quality score.”
RankSRGAN: Generative Adversarial Networks With Ranker for Image Super-Resolution (2019, methodology): “The Spearman Rank Order Correlation Coefficient (SROCC) [25] is a traditional evaluation metric to evaluate the performance of image quality assessment algorithms.”