A5038133707- Advanced Vision and Imaging
- Video Coding and Compression Technologies
- Advanced Image Processing Techniques
- Image and Signal Denoising Methods
- Advanced Data Compression Techniques
- Image and Video Quality Assessment
- Image Retrieval and Classification Techniques
- Advanced Image and Video Retrieval Techniques
- Image Enhancement Techniques
- Advanced Image Fusion Techniques
- Face recognition and analysis
- Biometric Identification and Security
- Video Analysis and Summarization
- Digital Filter Design and Implementation
- Advanced Neural Network Applications
- Generative Adversarial Networks and Image Synthesis
- Domain Adaptation and Few-Shot Learning
- Digital Media Forensic Detection
- Human Pose and Action Recognition
- Face and Expression Recognition
- Non-Invasive Vital Sign Monitoring
- Computer Graphics and Visualization Techniques
- Remote-Sensing Image Classification
- Visual Attention and Saliency Detection
- Speech and Audio Processing
City University of Hong Kong
2015-2024
Ben-Gurion University of the Negev
2020
ETH Zurich
2019
University of Hong Kong
2000-2008
South China University of Technology
2005
Hong Kong Chu Hai College
2005
Hong Kong Polytechnic University
1991-1994
Based on the real world image sequence's characteristic of center-biased motion vector distribution, a new four-step search (4SS) algorithm with checking point pattern for fast block estimation is proposed in this paper. A halfway-stop technique employed searching steps 2 to 4 and total number points varied from 17 27. Simulation results show that 4SS performs better than well-known three-step has similar performance (N3SS) terms compensation errors. In addition, also reduces worst-case...
This paper reviews the second challenge on spectral reconstruction from RGB images, i.e., recovery of whole- scene hyperspectral (HS) information a 3-channel image. As in previous challenge, two tracks were provided: (i) "Clean" track where HS images are estimated noise-free RGBs, themselves calculated numerically using ground-truth and supplied sensitivity functions (ii) "Real World" track, simulating capture by an uncalibrated unknown camera, recovered noisy JPEG-compressed images. A new,...
In block motion estimation, search patterns with different shapes or sizes and the center-biased characteristics of motion-vector distribution have a large impact on searching speed quality performance. We propose novel algorithm using cross-search pattern as initial step large/small diamond (DS) subsequent steps for fast estimation. The is designed to fit cross-center-biased vector real-world sequences by evaluating nine relatively higher probable candidates located horizontally vertically...
Remote imaging photoplethysmography (RIPPG) can achieve contactless monitoring of human vital signs. However, the robustness to a subject's motion is challenging problem for RIPPG, especially in facial video-based RIPPG. The RIPPG signal originates from radiant intensity variation skin with pulses blood and motions modulate skin. Based on optical properties skin, we build an model which origins artifacts be clearly described. region interest (ROI) regarded as Lambertian radiator effect ROI...
We propose two cross-diamond-hexagonal search (CDHS) algorithms, which differ from each other by their sizes of hexagonal patterns. These algorithms basically employ cross-shaped patterns consecutively in the very beginning steps and switch using diamond-shaped To further reduce checking points, pairs are proposed conjunction with candidates found located at diamond corners. Experimental results show that CDHSs perform faster than (DS) about 144% cross-diamond (CDS) 73%, whereas similar...
In personal healthcare, blood pressure (BP) is an important vital sign to be monitored frequently.However, traditional BP measurement devices require cuff's inflation and deflation that very uncomfortable for many users.Cuffless noninvasive estimation methods are attractive especially on using Photoplethysmography (PPG) approach achieving continuous monitoring minimal user's inconvenience.From recent studies the second derivative of PPG (SDPPG) vascular aging, SDPPG contains information...
Capturing visual image with a hyperspectral camera has been successfully applied to many areas due its narrowband imaging technology. Hyperspectral reconstruction from RGB images denotes reverse process of by discovering an inverse response function. Current works mainly map directly corresponding spectrum but do not consider context information explicitly. Moreover, the use encoder-decoder pair in current algorithms leads loss information. To address these problems, we propose 4-level...
In this paper, we propose an efficient general-purpose no-reference (NR) video quality assessment (VQA) framework that is based on 3D shearlet transform and convolutional neural network (CNN). Taking blocks as input, simple primary spatiotemporal features are extracted by transform, which capable of capturing natural scene statistics properties. Then, CNN logistic regression concatenated to exaggerate the discriminative parts predict a perceptual score. The resulting algorithm, name...
Many fast block-matching algorithms reduce computations by limiting the number of checking points. They can achieve high computation reduction, but often result in relatively higher matching error compared with full-search algorithm. A novel algorithm named normalized partial distortion search is proposed. The proposed reduces using a halfway-stop technique calculation block measure. In order to increase probability early rejection non-possible candidate motion vectors, accumulated and...
Fast block motion estimation normally consists of low-resolution coarse search and the following fine-resolution inner search. Most algorithms developed attempt to speed up without considering accelerating focused On top hexagonal method recently developed, an enhanced algorithm is proposed further improve performance in terms reducing number points distortion, where a novel fast employed by exploiting distortion information evaluated points. Our experimental results substantially justify...
Objective quality assessment has been widely used in image processing for decades and many researchers have studying the objective method based on Human Visual System (HVS). Recently Structural Similarity (SSIM) is proposed, under assumption that HVS highly adapted extracting structural information from a scene, simulation results proved it better than PSNR (or MSE). By deeply SSIM, we find fails measuring badly blurred images. Based this, develop an improved which called Edge-based (ESSIM)....
The state-of-the-art general-purpose no-reference image or video quality assessment (NR-I/VQA) algorithms usually rely on elaborated hand-crafted features which capture the Natural Scene Statistics (NSS) properties. However, designing these is not an easy problem. In this paper, we describe a novel NR-IQA framework based deep Convolutional Neural Networks (CNN). Directly taking raw as input and outputting score, new integrates feature learning regression into one optimization process,...
Given a grayscale photograph, the colorization system estimates visually plausible colorful image. Conventional methods often use semantics to colorize images. However, in these methods, only classification semantic information is embedded, resulting confusion and color bleeding final colorized To address issues, we propose fully automatic Saliency Map-guided Colorization with Generative Adversarial Network (SCGAN) framework. It jointly predicts saliency map minimize Since global features...
Deep convolutional neural networks (CNNs) have been successfully applied on no-reference image quality assessment (NR-IQA) with respect to human perception. Most of these methods deal small patches and use the average score test for predicting whole quality. We discovered that from homogenous regions are unreliable both network training final estimation. In addition, complex structures much higher chances achieving better prediction. Based findings, we enhanced conventional CNN-based NR-IQA...
This paper reviews the first AIM challenge on mapping camera RAW to RGB images with focus proposed solutions and results. The participating teams were solving a real-world photo enhancement problem, where goal was map original low-quality from Huawei P20 device same photos captured Canon 5D DSLR camera. considered problem embraced number of computer vision subtasks, such as image demosaicing, denoising, gamma correction, resolution sharpness enhancement, etc. target metric used in this...
We propose a hybrid recurrent Video Colorization with Hybrid Generative Adversarial Network (VCGAN), an improved approach to video colorization using end-to-end learning. The VCGAN addresses two prevalent issues in the domain: Temporal consistency and unification of network refinement into single architecture. To enhance quality spatiotemporal consistency, mainstream generator is assisted by additional networks, i.e., global feature extractor placeholder extractor, respectively. encodes...
In most block-based video coding systems, the fast block matching algorithms (BMAs) use origin as initial search center, which may not track motion very well. To improve accuracy of BMAs, a new adaptive tracking algorithm is proposed. Based on spatial correlation blocks, predicted starting point, reflects trend current block, adaptively chosen. This center found closer to global minimum, and thus center-biased BMAs can be used find vector more efficiently. Experimental results show that...
The quality control for video coding usually absents from many traditional fast block motion estimators. A novel block-matching algorithm estimation named the adjustable partial distortion search (APDS) is proposed. It a new normalized comparison method capable of adjusting prediction accuracy against searching speed by factor k. With adjustability, APDS could act as (NPDS) when k equal to 0, and conventional (PDS) 1. In addition, it uses halfway-stop technique with progressive distortions...