Article
Engineering, Electrical & Electronic
Ajay Kumar Reddy Poreddy, Raja Bharath Chandra Ganeswaram, Balasubramanyam Appina, Priyanka Kokil, Ram Bilas Pachori
Summary: In this article, a no-reference virtual reality image quality assessment (VR IQA) model based on global and local natural scene statistics (NSS) is proposed. The model utilizes generalized Gaussian distributions (GGDs) to compute global features and statistical properties of spatial and spectral entropy maps for local features. Support vector regressor (SVR) is employed to map the quality-aware feature set to VR image quality. Experimental results demonstrate state-of-the-art performance compared to existing 2D and 3D IQA models.
IEEE TRANSACTIONS ON INSTRUMENTATION AND MEASUREMENT
(2023)
Article
Engineering, Electrical & Electronic
Weihao Xia, Yujiu Yang, Jing-Hao Xue, Jing Xiao
Summary: This study introduces a new no-reference image quality assessment method, incorporating the concept of domain fingerprint. By designing a new domain-aware architecture, the method is able to simultaneously determine the distortion sources and quality of an image. Experimental results show that the proposed method outperforms most existing state-of-the-art NR-IQA methods.
IEEE TRANSACTIONS ON CIRCUITS AND SYSTEMS FOR VIDEO TECHNOLOGY
(2021)
Article
Neurosciences
Yuhong Wang, Hong Li, Qiuping Jiang
Summary: Omnidirectional images (ODIs) have gained significant attention in virtual reality (VR) due to their ability to provide an immersive experience. However, ODIs often suffer from quality degradations during processing, making quality assessment crucial in the VR community. In this article, a novel no-reference (NR) ODI quality evaluation method is proposed, which focuses on constructing a dynamically attentive viewport sequence (DAVS) and extracting quality-aware features (QAFs). Experimental results demonstrate that the proposed method achieves state-of-the-art performance.
FRONTIERS IN NEUROSCIENCE
(2022)
Article
Engineering, Electrical & Electronic
Milosz Rajchel, Mariusz Oszust
Summary: The study introduces a new benchmark database and a new NR-IQA metric that uses a wide range of image features to address various distortions, achieving superior correlation results with human scores compared to state-of-the-art IQA methods.
SIGNAL IMAGE AND VIDEO PROCESSING
(2021)
Article
Computer Science, Artificial Intelligence
Jie Song, Mengjun Liu
Summary: This paper proposes a new methodology to convert a full-reference focus quality assessment metric into a no-reference one. The methodology includes three hypotheses that describe the relationship between focus quality of the original image and its variants. Two no-reference metrics are constructed using this methodology, one using Brenner Gradient and the other using a full-reference metric proposed by the authors. Evaluation is conducted on both a public dataset and a proposed dataset, showing that the second metric exhibits the best performance with comparable calculation time to some fastest metrics considered. (c) 2023 Elsevier Ltd. All rights reserved.
PATTERN RECOGNITION
(2023)
Article
Chemistry, Analytical
Liquan Shen, Yang Yao, Xianqiu Geng, Ruigang Fang, Dapeng Wu
Summary: This paper proposes a novel no-reference quality assessment metric for stereoscopic images using natural scene statistics and 3D visual perceptual information. Features of the stereoscopic images are extracted based on the natural scene statistics model, and the binocular rivalry effect and other 3D visual properties are also considered. Experimental results show that the proposed method achieves good alignment with subjective assessment of stereoscopic images.
Article
Computer Science, Artificial Intelligence
Baoliang Chen, Haoliang Li, Hongfei Fan, Shiqi Wang
Summary: In this paper, a novel unsupervised domain adaptation based no-reference quality assessment method is developed to transfer the quality characteristics of natural images to images captured by non-optical cameras. By introducing three complementary losses, the method achieves higher performance through a progressive regularization of the feature space of ranking.
IEEE TRANSACTIONS ON IMAGE PROCESSING
(2021)
Article
Computer Science, Artificial Intelligence
Baoliang Chen, Lingyu Zhu, Chenqi Kong, Hanwei Zhu, Shiqi Wang, Zhu Li
Summary: In this paper, a no-reference image quality assessment method based on feature level pseudo-reference hallucination is proposed. The method utilizes perceptually meaningful features to characterize visual quality and leverages natural image statistical behaviors for accurate predictions. Experimental results demonstrate the effectiveness and high generalization capability of the proposed method on multiple databases.
IEEE TRANSACTIONS ON IMAGE PROCESSING
(2022)
Article
Engineering, Electrical & Electronic
Neeraj Badal, Rajiv Soundararajan, Ankur Garg, Abhishek Patil
Summary: Pansharpening is a process that enhances the spatial resolution of a multispectral image using a high-resolution panchromatic image. Quality assessment of pansharpened images is crucial for the analysis and design of pansharpening methods. This article focuses on predicting the quality in a no-reference setting and proposes a learning-based approach for image quality assessment.
IEEE JOURNAL OF SELECTED TOPICS IN APPLIED EARTH OBSERVATIONS AND REMOTE SENSING
(2022)
Article
Computer Science, Artificial Intelligence
Juan Wang, Zewen Chen, Chunfeng Yuan, Bing Li, Wentao Ma, Weiming Hu
Summary: This paper proposes a hierarchical curriculum learning (HCL) framework for no-reference image quality assessment (NR-IQA). The framework leverages external data to learn prior knowledge about IQA and achieves state-of-the-art performance on multiple authentic IQA datasets.
INTERNATIONAL JOURNAL OF COMPUTER VISION
(2023)
Article
Computer Science, Information Systems
Zahi Al Chami, Chady Abou Jaoude, Richard Chbeir, Mahmoud Barhamgi, Mansour Naser Alraja
Summary: This paper presents a framework for processing a large number of images, which can real-time estimate and assist in enhancing image quality. The quality evaluation is conducted using Convolutional Neural Network and other methods to achieve both No-Reference and Full-Reference quality assessment, enhanced with a Super-Resolution Model.
MULTIMEDIA TOOLS AND APPLICATIONS
(2022)
Article
Engineering, Electrical & Electronic
Kamal Lamichhane, Marco Carli, Federica Battisti
Summary: Assessing image quality is challenging, requiring evaluation by subjects or defining an objective quality metric. This study proposes an objective quality metric based on a deep neural network, considering the human vision system by computing saliency maps and natural scene statistics features. The metric consists of convolutional layers and regression units, with the first module trained using preprocessed distorted images and feature weights smoothed by the estimated saliency map. The second module is adjusted using ground truth quality scores and scaled feature weights obtained from the first module. The proposed metric's performance was evaluated on four datasets, showing effective matching with ground truth scores.
SIGNAL PROCESSING-IMAGE COMMUNICATION
(2023)
Article
Computer Science, Hardware & Architecture
Tian Yuan, Chen Li, Lihua Tian, Guo Li
Summary: Recent years have seen significant progress in image quality assessment, especially in the field of no-reference (NR)-IQA with the development of deep learning. The proposed framework in this study utilizes a range mapping method to enhance the accuracy and generalization of NR-IQA models by mapping existing full-reference (FR)-IQA datasets to NR-IQA datasets. Experimental results using the largest available datasets have confirmed the effectiveness of this approach.
COMPUTERS & ELECTRICAL ENGINEERING
(2021)
Article
Engineering, Electrical & Electronic
Zicheng Zhang, Wei Sun, Xiongkuo Min, Tao Wang, Wei Lu, Guangtao Zhai
Summary: To improve the viewer's Quality of Experience (QoE) and optimize computer graphics applications, 3D model quality assessment (3D-QA) has become an important task in the multimedia area. In this paper, a no-reference (NR) quality assessment metric for colored 3D models represented by both point cloud and mesh is proposed. The method projects the 3D models into quality-related domains, extracts quality-aware features using 3D natural scene statistics (3D-NSS) and entropy, and employs a support vector regression (SVR) model to predict the visual quality scores. Experimental results show that the proposed method outperforms most compared NR 3D-QA metrics and reduces the performance gap with state-of-the-art FR 3D-QA metrics.
IEEE TRANSACTIONS ON CIRCUITS AND SYSTEMS FOR VIDEO TECHNOLOGY
(2022)
Article
Computer Science, Information Systems
Guiying Zhu
Summary: In this research, we propose an efficient non-reference image quality assessment method called TDSDQA, which utilizes natural scene statistics (NSS) in the nonsubsampled contourlet transform (NSCT) domain and spatial domain. By analyzing the correlation of coefficients in the NSCT domain and calculating their mutual information, we describe their correlation. The structure similarity of coefficients is then used to represent picture structure information statistics. Additionally, 84 statistics features are extracted to predict image quality scores using support vector regression (SVR) approach, and the method is tested on LIVE and TID2008 IQA databases. Experimental results demonstrate that this method is suitable for various image distortion types and performs competitively compared to other state-of-the-art NR-IQA algorithms.
JOURNAL OF INFORMATION SCIENCE AND ENGINEERING
(2023)
