- Image and Video Quality Assessment
- Advanced Image Processing Techniques
- Visual Attention and Saliency Detection
- Advanced Vision and Imaging
- Advanced Image Fusion Techniques
- Image Enhancement Techniques
- Video Coding and Compression Technologies
- Advanced Optical Imaging Technologies
- Advanced Computing and Algorithms
- Virtual Reality Applications and Impacts
- Data Visualization and Analytics
- Visual perception and processing mechanisms
- Multimedia Communication and Technology
- Image Processing Techniques and Applications
- Literary Theory and Cultural Hermeneutics
- Topic Modeling
- Adversarial Robustness in Machine Learning
- Advanced Text Analysis Techniques
- American and British Literature Analysis
- Mobile Crowdsensing and Crowdsourcing
- Video Surveillance and Tracking Methods
- Advanced Neural Network Applications
- Image and Signal Denoising Methods
- Forensic Fingerprint Detection Methods
- Complex Network Analysis Techniques
Alibaba Group (China)
2019-2024
Zaozhuang University
2022-2024
Hong Kong Polytechnic University
2023
Fraunhofer Institute for Telecommunications, Heinrich Hertz Institute
2023
Google (United States)
2023
Texas State University
2023
Institut Universitaire de France
2023
Jiangsu University
2022
Alibaba Group (United States)
2022
Chengdu University
2021
Scatterplots and parallel coordinate plots (PCPs) that can both be used to assess correlation visually. In this paper, we compare these two visualization methods in a controlled user experiment. More specifically, 25 participants were asked report observed as function of the sample under varying conditions method, size observation time. A statistical model is proposed describe judgment process. The accuracy bias judgments different are established by interpreting parameters model....
Recently, many methods have been proposed to predict the image quality which is generally described by mean opinion score (MOS) of all subjective ratings given an image. However, few efforts focus on predicting distribution ratings. In fact, reflecting diversity, uncertainty, etc., can provide more information about than a single MOS, worthy in-depth study. this paper, we propose convolutional neural network based fuzzy theory quality. The method consists three main steps: feature...
With the fast proliferation of online video sites and social media platforms, user, professionally occupationally generated content (UGC, PGC, OGC) videos are streamed explosively shared over Internet. Consequently, it is urgent to monitor quality these Internet guarantee user experience. However, most existing modern assessment (VQA) databases only include UGC cannot meet demands for other kinds with real-world distortions. To this end, we collect 1,072 from Youku, a leading Chinese hosting...
Free viewpoint videos (FVVs) provide immersive experiences for end-users, and they have been applied in many applications, such as movies, sports, TV shows. However, the development of quantifying quality experience (QoE) FVVs is still relatively slow due to high costs data collection limited public databases. In this paper, we conduct a comprehensive study on FVV QoE. First, construct largest, best our knowledge, QoE database called Youku-FVV from two complex real scenarios, i. e.,...
We present GSO-Simulcast, a new architecture designed for large-scale multi-party video-conferencing systems. GSO-Simulcast is currently deployed at full-scale in Alibaba's Dingtalk video conferencing that serves more than 500 million users. It marks fundamental shift from today's Simulcast, where media server locally decides how to switch and forward streams based on fragmented network view. Instead, globally orchestrates the publishing, subscribing, as well resolution bitrate of each...
Pill image recognition is vital for many personal/public health-care applications and should be robust to diverse unconstrained real-world conditions. Most existing pill models are limited in tackling this challenging few-shot learning problem due the insufficient instances per category. With training data, neural network-based have limitations discovering most discriminating features, or going deeper. Especially, fail handle hard samples taken under less controlled imaging In study, a new...
Saliency detection is an effective front-end process to many security-related tasks, <italic xmlns:mml="http://www.w3.org/1998/Math/MathML" xmlns:xlink="http://www.w3.org/1999/xlink">e.g.</i> automatic drive and tracking. Adversarial attack serves as efficient surrogate evaluate the robustness of deep saliency models before they are deployed in real world. However, most current adversarial attacks exploit gradients spanning entire image space craft examples, ignoring fact that natural images...
Deep convolutional neural networks (CNNs) have increasingly become a prominent method for blind image quality assessment (BIQA). The process of typically involves feature extraction, average-based pooling, and regression. Based on this process, as well the consensus that visual an mainly relies its content distortions, work improves CNNs BIQA in two ways. First, considering content-awareness perception, we incorporate via dynamic filtering module to extract content-adaptive features...
User expectations have a crucial impact on the levels of quality experience (QoE) that they consider acceptable or satisfying. Measuring acceptability and annoyance has mainly been performed in separate multi-step experiments without any control over participants' expectations. This paper introduces simple methodology to obtain information about both entities single step compares several data processing strategies useful for results interpretation. A specifically designed subjective...
Deep neural networks are vulnerable to adversarial attacks. More importantly, some examples crafted against an ensemble of source models transfer other target and, thus, pose a security threat black-box applications (when attackers have no access the models). Current transfer-based attacks, however, only consider limited number craft example obtain poor transferability. Besides, recent query-based which require numerous queries model, not come under suspicion by model but also cause...
Accurate measurement of perceptual quality is important for various immersive multimedia, which demand real-time control or quality-based bench-marking relevant algorithms. For instance, virtual views rendering in Free-Viewpoint (FV) navigation scenarios a typical case that introduces challenging distortions, particularly the ones around dis-occluded regions. Existing metrics, most are targeting impairments caused by compression network condition, fail to quantify such non-uniform...
The swift development of the multimedia technology has raised dramatically users' expectation on quality experience. To obtain ground-truth perceptual for model training, subjective assessment is necessary. Crowdsourcing platform provides us a convenient and feasible way to run large-scale experiments. However, obtained labels are generally noisy. In this paper, we propose probabilistic graphical annotation infer underlying ground truth discovering annotator's behavior. proposed model, label...
Virtual viewpoints synthesis is an essential process for many immersive applications including Free-viewpoint TV (FTV). A widely used technique Depth-Image-Based-Rendering (DIBR) technique. However, such may introduce challenging non-uniform spatial-temporal structure-related distortions. Most of the existing state-of-the-art quality metrics fail to handle these distortions, especially temporal structure inconsistencies observed during switch different viewpoints. To tackle this problem,...
We present a large-scale eye tracking database for stereo-scopic video. A set of participants were involved in this experiment. The human fixation maps created as the ground truth stereoscopic video from gaze data participants. To best our knowledge, is first visual attention modeling details processing operations and properties are described paper.
One key in image quality assessment (IQA) is the design of representations that can capture changes structures caused by distortions. Recent studies show sparse coding has emerged as a promising approach to analyzing for IQA. However, existing sparse-coding-based IQA approaches use linear models, which ignore nonlinearities manifolds patches and thus cannot analyze complex well. To overcome such weakness, this paper, we introduce nonlinear A kernel dictionary construction scheme proposed,...
Subjective assessment of Quality Experience in stereoscopic 3D requires new guidelines for the environmental setup as existing standards such ITU-R BT.500 may no longer be appropriate. A first step is to perform cross-lab experiments different viewing conditions on same video sequences. Three international labs performed Absolute Category Rating studies a freely available database containing degradations that are mainly related quality degradations. Different have been used labs: Passive...
Millions of users are active on social media. To allow to better showcase themselves and network with others, we explore the auto-generation media self-introduction, a short sentence outlining user’s personal interests. While most prior work profiling tags (e.g., ages), investigate sentence-level self-introductions provide more natural engaging way for know each other. Here exploit tweeting history generate their self-introduction. The task is non-trivial because content may be lengthy,...
In this paper, we proposed a no-reference (NR) quality metric for RGB plus image-depth (RGB-D) synthesis images based on Generative Adversarial Networks (GANs), namely GANs-NQM. Due to the failure of inpainting dis-occluded regions in RGB-D process, capture non-uniformly distributed local distortions and learn their impact perceptual are challenging tasks objective metrics. our study, characteristics GANs, i) novel training strategy GANs using existing large-scale computer vision datasets...
The Internet streaming is changing the way of watching videos for people. Traditional quality assessment on cable/satellite broadcasting system mainly focused perceptual quality. Nowadays, this concept has been extended to Quality Experience (QoE) which considers also contextual factors, such as environment, display devices, etc. In study, we focus influence devices QoE. A subjective experiment was conducted by using our proposed AccAnn methodology. observers evaluated QoE video sequences...
The development of rigorous quality assessment model relies on the collection reliable subjective data, where perceived visual multimedia is rated by human observers. Different protocols can be used according to objectives, which determine discriminability and accuracy data. Single stimulus methodology, e.g., Absolute Category Rating (ACR) has been widely adopted due its simplicity efficiency. However, Pair Comparison (PC) significant advantage over ACR in terms discriminability. In...
As the immersive multimedia techniques like Free-viewpoint TV (FTV) develop at an astonishing rate, user's demand for high-quality contents increases dramatically. Unlike traditional uniform artifacts, distortions within could be non-uniform structure-related and thus are challenging commonly used quality metrics. Recent studies have demonstrated that representation of visual features can extracted from multiple levels hierarchy. Inspired by hierarchical mechanism in human system (HVS), this...