MetaIQA: Deep Meta-Learning for No-Reference Image Quality Assessment

Zhu, Hancheng; Li, Leida; Wu, Jinjian; Dong, Weisheng; Shi, Guangming

doi:10.1109/cvpr42600.2020.01415

Cited by 306 publications

(131 citation statements)

References 42 publications

Supporting

Mentioning

129

Contrasting

Order By: Relevance

“…Low-quality image classification can be a challenging task. In the future, more advanced techniques or operators, such as image quality assessment method [62] and data argumentation, can be investigated to improve the performance of low-quality image classification tasks.…”

Section: Discussionmentioning

confidence: 99%

Genetic Programming-Based Discriminative Feature Learning for Low-Quality Image Classification

Bi¹,

Xue²,

Zhang³

2021

Preprint

View full text Add to dashboard Cite

Being able to learn discriminative features from low-quality images has raised much attention recently due to their wide applications ranging from autonomous driving to safety surveillance. However, this task is difficult due to high variations across images, such as scale, rotation, illumination, and viewpoint, and distortions in images, such as blur, low contrast, and noise. Image preprocessing could improve the quality of the images, but it often requires human intervention and domain knowledge. Genetic programming (GP) with a flexible representation can automatically perform image preprocessing and feature extraction without human intervention. Therefore, this study proposes a new evolutionary learning approach using GP (EFLGP) to learn discriminative features from images with blur, low contrast, and noise for classification. In the proposed approach, we develop a new program structure (individual representation), a new function set, and a new terminal set. With these new designs, EFLGP can detect small regions from a large input low-quality image, select image operators to process the regions or detect features from the small regions, and output a flexible number of discriminative features. A set of commonly used image preprocessing operators is employed as functions in EFLGP to allow it to search for solutions that can effectively handle low-quality image data. The performance of EFLGP is comprehensively investigated on eight datasets of varying difficulty under the original (clean), blur, low contrast, and noise scenarios, and compared with a large number of benchmark methods using handcrafted features and deep features. The experimental results show that EFLGP achieves significantly better or similar results in most comparisons. The results also reveal that EFLGP is more invariant than the benchmark methods to blur, low contrast, and noise.

show abstract

Section: Discussionmentioning

confidence: 99%

Genetic Programming-Based Discriminative Feature Learning for Low-Quality Image Classification

Bi¹,

Xue²,

Zhang³

2021

Preprint

View full text Add to dashboard Cite

show abstract

“…[13] utilized GAN to model an active inference module, and image quality is measured based on its primary content from such module. [14] first apply meta-learning into the field of IQA, the prior knowledge is collected by a meta-network, and they adapt the pre-train knowledge to the specific domain via small-sample fine-tuning. [15] employed hyper network to generated an exclusive representation for each picture, which achieves surprise results by mapping this representation to quality score latent space.…”

Section: Image Quality Assessment Methodsmentioning

confidence: 99%

Blind Image Quality Assessment for a Single Image From Text-to-Image Synthesis

Zhang

et al. 2021

IEEE Access

View full text Add to dashboard Cite

A fundamental bottleneck in text-to-image synthesis is that there are rarely subjective quality evaluation metrics for a single generated image. To address this issue, this paper proposed a procedure to evaluate the single generated image, which includes a specific dataset named multiple metrics quality assessment for birds(MMQA Birds) and a learning model named blind generated image evaluator(BGIE). The motivation of our proposal is twofold. On the one hand, subjective image quality evaluation is a human perceptual task; Therefore, it tends to be a process of supervised learning. To the best of our knowledge, there are not any datasets for this study. Thus, we handle this problem via designing a specific dataset. On the other hand, we observed that the spatial content of generated image attracts more attention when humans judge its quality; According to this finding, an efficient machine-learning model that combines both pixellevel features and spatial features is proposed. Extensive experiments manifest our method can solve this problem to some extent. In the generated image dataset, BGIE surpasses the state-of-art NSS-based method by 6.3% in PLCC and SRCC. In practice, we further discuss the rationality of the MMQA Birds dataset and the application of BGIE. It proves that both in subjective and objective aspects, our method achieves convincing results.

show abstract

“…For example, Kang et al [7][8] proposed a multi-task shallow CNN to learn both the distortion type and the quality score; Kim and Lee [9] applied state-of-the-art FR-IQA methods to provide proxy quality scores for each image patch as the ground truth label in the pre-training stage, and the proposed network was fine-tuned by the Subjective annotations. Similarly, Da Pan et al [10] employed the U-Net to learn the local quality predicting scores previously calculated by Full-Reference IQA methods, several Dense layers were then incorporated to pool the local quality predicting scores into an overall perceptual quality score; Liang et al [11] tried to utilize similar scene as reference to provide more prior information for the IQA model; Liu et al [12] proposed to use RankNet to learn the quality rank information of image pairs in the training set, and then used the output of the second last layer to predict the quality score; Yee et al [13] tried to learn the corresponding unknown reference image from the distorted one by resorting the Generative Adversarial Networks, and to assess the perceptual quality by comparing the hallucinated reference image and the distorted image; Chiu et al [1] proposed a new IQA framework and corresponding dataset that links the IQA issue to two practical vision tasks which are image captioning and visual question answering respectively; Su et al [14] employed self-adaptive hyper network whose parameters could adjust according to image contents; Zhu et al [15] leveraged meta-learning to learn a general-purpose BIQA model from training set of several specific distortion types.…”

Section: Related Workmentioning

confidence: 99%

Group Perceptual Quality Optimization for Multi-Channel Image Encoding Systems Based on Adaptive Hyper Networks

Chen

Chen³

et al. 2021

IEEE Access

View full text Add to dashboard Cite

Images and short videos that produced by social networks surge in recent years. Image/Video encoders, such as JPEG and H.264, are indispensably involved to reduce the transmitting bandwidth. However, based on our observation, the encoding parameters and their candidates are often preset to fixed values (or fixed candidate values) in real-world scenarios, which might not be the optimal bandwidth allocation strategy. Considering that, we propose an efficient group quality optimization (GQO) framework for multichannel image/video encoding systems in which the encoding parameters are configured in a perceptualquality-driven manner. The GQO framework employs adaptive hyper network to predict the relationships between encoding parameters, transmitting resources, and perceptual qualities, i.e., just taking the pristine image as input, the adaptive hyper network could accurately yield a global overview of perceptual quality and transmitting resource varied along encoding parameters. A step-by-step optimization procedure is then employed to search the optimal encoding parameter for each channel so that overall perceptual quality could be maximized under limited transmitting resource. Experimental results demonstrate the proposed GQO framework could achieve higher perceptual quality whilst maintain the same bandwidth compared to traditional allocation strategies where encoding parameters are preset.

show abstract

MetaIQA: Deep Meta-Learning for No-Reference Image Quality Assessment

Cited by 306 publications

References 42 publications

Genetic Programming-Based Discriminative Feature Learning for Low-Quality Image Classification

Genetic Programming-Based Discriminative Feature Learning for Low-Quality Image Classification

Blind Image Quality Assessment for a Single Image From Text-to-Image Synthesis

Group Perceptual Quality Optimization for Multi-Channel Image Encoding Systems Based on Adaptive Hyper Networks

Contact Info

Product

Resources

About