KonIQ-10k: Towards an ecologically valid and large-scale IQA database

Lin, Hanhe; Hosu, Vlad; Saupe, Dietmar

doi:10.48550/arxiv.1803.08489

Cited by 19 publications

(85 citation statements)

References 13 publications

Supporting

Mentioning

Contrasting

Order By: Relevance

“…For target datasets, we choose two datasets with authentic distortion (KonIQ-10k [4] and LIVE Challenge (LIVEC) [20]) and two datasets with synthetic distortion (LIVE [5] and CSIQ [42]). KonIQ-10k consists of 10073 images which are selected from the large public multimedia database YFCC100m [43].…”

Section: Methodsmentioning

confidence: 99%

“…a) Archtecture. : In this section, we provide experiments of network architecture on KonIQ dataset [4], which is shown in Figure 5 and the results are shown in Table IX. All the results are test on models trained on 150000 epochs, and all the other parameters are kept consistent.…”

Section: Ablation Studymentioning

confidence: 99%

“…It is observed that when Node Builder with 3 fully connected layers, Edge Builder with 3 graph convolutional layers and TDN with 3 graph convolutional layers, our model achieve the best performance. Table X shows the relationship between the number of epoch and performance on KonIQ dataset [4]. As the training progresses, the model's ability to distinguish and represent each distortion is improved, so that the performance XII shows that when the size of node embedding is set as 256, the performance is the best.…”

Section: Ablation Studymentioning

confidence: 99%

“…The FPN extracts the distributional characteristics of the samples in DGR, and predicts fuzzy levels based on a Gaussian prior since the subjective quality ratings are often biased by image content. In the end, as the learned DGRs can model the relationship between perceptual image quality and distortion-related factors, GraphIQA can achieve the state-of-the-art on most typical IQA datasets (e.g., KonIQ-10k [4] or LIVE [5]). Note that the training cost of DGRs is very low as it only requires weak supervision, i.e.…”

Section: Introductionmentioning

confidence: 99%

See 3 more Smart Citations

GraphIQA:Learning Distortion Graph Representations for Blind Image Quality Assessment

Sun¹,

Chen²,

Xu³

et al. 2021

Preprint

View full text Add to dashboard Cite

Learning-based blind image quality assessment (BIQA) methods have recently drawn much attention for their superior performance compared to traditional methods. However, most of them do not effectively leverage the relationship between distortion-related factors, showing limited feature representation capacity. In this paper, we show that human perceptual quality is highly correlated with distortion type and degree, and is biased by image content in most IQA systems. Based on this observation, we propose a Distortion Graph Representation (DGR) learning framework for IQA, called GraphIQA. In GraphIQA, each distortion is represented as a graph i.e. DGR. One can distinguish distortion types by comparing different DGRs, and predict image quality by learning the relationship between different distortion degrees in DGR. Specifically, we develop two sub-networks to learn the DGRs: a) Type Discrimination Network (TDN) that embeds DGR into a compact code for better discriminating distortion types; b) Fuzzy Prediction Network (FPN) that extracts the distributional characteristics of the samples in DGR and predicts fuzzy degrees based on a Gaussian prior. Experiments show that our GraphIQA achieves the stateof-the-art performance on many benchmark datasets of both synthetic and authentic distortion. The code is available at https://github.com/geekyutao/GraphIQA.

show abstract

Section: Methodsmentioning

confidence: 99%

Section: Ablation Studymentioning

confidence: 99%

Section: Ablation Studymentioning

confidence: 99%

Section: Introductionmentioning

confidence: 99%

See 2 more Smart Citations

GraphIQA:Learning Distortion Graph Representations for Blind Image Quality Assessment

Sun¹,

Chen²,

Xu³

et al. 2021

Preprint

View full text Add to dashboard Cite

show abstract

“…Recently, convolutional neural networks (CNN) have been shown to deliver remarkable performance on a wide range of computer vision tasks [23]- [25]. Several deep CNN-based BVQA models have also been proposed [18], [19], [26], [27] by training them on recently created large-scale psychometric databases [28], [29]. These methods have yielded promising results on synthetic distortion datasets [1], but still struggled on UGC quality assessment databases [30]- [32].…”

Section: Introductionmentioning

confidence: 99%

RAPIQUE: Rapid and Accurate Video Quality Prediction of User Generated Content

Tu,

Yu,

Wang

et al. 2021

Preprint

View full text Add to dashboard Cite

Blind or no-reference video quality assessment of user-generated content (UGC) has become a trending, challenging, unsolved problem. Accurate and efficient video quality predictors suitable for this content are thus in great demand to achieve more intelligent analysis and processing of UGC videos. Previous studies have shown that natural scene statistics and deep learning features are both sufficient to capture spatial distortions, which contribute to a significant aspect of UGC video quality issues. However, these models are either incapable or inefficient for predicting the quality of complex and diverse UGC videos in practical applications. Here we introduce an effective and efficient video quality model for UGC content, which we dub the Rapid and Accurate Video Quality Evaluator (RAPIQUE), which we show performs comparably to state-of-the-art (SOTA) models but with orders-of-magnitude faster runtime. RAPIQUE combines and leverages the advantages of both quality-aware scene statistics features and semantics-aware deep convolutional features, allowing us to design the first general and efficient spatial and temporal (space-time) bandpass statistics model for video quality modeling. Our experimental results on recent large-scale UGC video quality databases show that RAPIQUE delivers top performances on all the datasets at a considerably lower computational expense. We hope this work promotes and inspires further efforts towards practical modeling of video quality problems for potential real-time and low-latency applications. To promote public usage, an implementation of RAPIQUE has been made freely available online: https://github.com/vztu/RAPIQUE.

show abstract

No reference image quality assessment using gabor convolutional neural networks

Vadlamudi,

Khan Md

2024

Eng. Res. Express

View full text Add to dashboard Cite

With the emergence of image capturing devices and increase in usage of internet, massive volume of network data was occupied with digital images. For efficient data transmission, the image may undergo several processing units in from the point it captures to the display/storage device. It may result in loss of the perceptual image quality. Therefore, it is necessary to estimate the image quality to measure the quality of experience. It was found that the convolutional neural networks serve as potential tool for effective feature extraction in many image processing applications. Particularly, with the first layer as Gabor filters, the robustness of the network can be reinforced with learnable Gabor parameters. This paper proposes Gabor Convolutional Neural Network method for No-Reference Image Quality Assessment. Their well-defined spatial structured filters are promising in extracting quality features from the local patches and maps them to perceptual quality scores. Our proposed architecture was tested over synthetic and authentic databases such as LIVE, TID2013, CSIQ, LIVE-MD, MDID2016, LIVE Wild and KoNiQ-10k. The proposed approach was also tested on the Waterloo 3D phase-II database, which contains high-resolution images of both the eyes individually with their respective DMOS scores. The proposed approach out performs over LIVE-MD and LIVE Wild and competes with existing algorithm over other databases.

show abstract

KonIQ-10k: Towards an ecologically valid and large-scale IQA database

Cited by 19 publications

References 13 publications

GraphIQA:Learning Distortion Graph Representations for Blind Image Quality Assessment

GraphIQA:Learning Distortion Graph Representations for Blind Image Quality Assessment

RAPIQUE: Rapid and Accurate Video Quality Prediction of User Generated Content

No reference image quality assessment using gabor convolutional neural networks

Contact Info

Product

Resources

About