A5025083064- Advanced Vision and Imaging
- Image Processing Techniques and Applications
- Advanced Image Processing Techniques
- Image and Signal Denoising Methods
- Advanced Data Compression Techniques
- Optical measurement and interference techniques
- Advanced Optical Sensing Technologies
- Biometric Identification and Security
- Advanced Optical Imaging Technologies
- Statistical Methods and Inference
- CCD and CMOS Imaging Sensors
- Video Coding and Compression Technologies
- Video Surveillance and Tracking Methods
- Infrared Target Detection Methodologies
- Statistical Methods in Clinical Trials
- User Authentication and Security Systems
- Advanced Fluorescence Microscopy Techniques
- Image and Video Stabilization
- Hand Gesture Recognition Systems
- Industrial Vision Systems and Defect Detection
- Face recognition and analysis
- Video Analysis and Summarization
- Image Enhancement Techniques
- Health Systems, Economic Evaluations, Quality of Life
- Photoacoustic and Ultrasonic Imaging
Xi'an Jiaotong University
2013-2024
Zhejiang University
2020-2022
Institute of Art
2009
Perceptual image compression has shown strong potential for producing visually appealing results at low bitrates, surpassing classical standards and pixel-wise distortion-oriented neural methods. However, existing methods typically improve performance by incorporating explicit semantic priors, such as segmentation maps textual features, into the encoder or decoder, which increases model complexity adding parameters floating-point operations. This limits model's practicality, often occurs on...
The goal of image rescaling is to embed the information from high-resolution images into low-resolution and then reconstruct in reverse. Existing methods either focus on small scaling factors or do not generalize well natural with diverse content extreme settings, i.e., using (e.g., 16× 32×). When performing rescaling, previous often fail produce plausible high-quality results due insufficient cues images. In this work, we propose an framework that exploits rich generative prior integrated...
Abstract Many learning‐based approaches to image deblurring have received increasing attention in recent years. However, the models trained on existing synthetic datasets do not generalize well real‐world blur, resulting undesirable artifacts and residual blur. This work attempts address this problem from two aspects: training data synthesis network architecture. To narrow domain gap between real domains, a realistic blur pipeline generate high‐quality blurred is proposed. Since non‐uniform...
Since the rapid development of face recognition systems using 3D cameras, public has demanded great safety regulations for these devices. As a closely related topic, multimodal anti-spoofing (FAS) become an indispensable part systems. However, existing FAS tools suffer from performance degradation under external low-lighting conditions and insufficient representation capabilities fusion features. To address issues, we present attention-aware dual-stream method cameras (i.e., IR+Depth)...
Depth information has been used in many fields because of its low cost and easy availability, since the Microsoft Kinect was released. However, Kinect-like RGB-D sensors show limited performance certain applications place high demands on accuracy robustness depth information. In this paper, we propose a sensing system that contains laser projector similar to Kinect, two infrared cameras located both sides projector, obtain higher spatial resolution We apply block-matching algorithm estimate...
Under-display imaging has recently received considerable attention in both academia and industry. As a variation of this technique, under-display ToF (UD-ToF) cameras enable depth sensing for full-screen devices. However, it also brings problems image blurring, signal-to-noise ratio ranging accuracy reduction. To address these issues, we propose cascaded deep network to improve the quality UD-ToF maps. The comprises two subnets, with first using complex-valued raw domain perform denoising,...
most YOLO object detection neural networks prefer to focus on traditional RGB image, but previous studies rarely consider special network with compact architecture for infrared image. In this paper, we analyze original architecture, and propose a based by using different blocks from work small target We use layer, GhostConv convolution, Focus structure, Focal EIOU Loss soft NMS modules improve which improves the accuracy speed of The experimental results show that model can be effectively...
This paper presents an efficient motion adaptive de-interlacing technique that consists of two main steps, i.e. 4-field extended Gaussian filtering detection and adjustable window ELA de-interlacing. Four consecutive interlaced fields are used to detect accurately. With a filter, the can eliminate influence noise. An is adopt process pixels, which reconstruct image with high quality even in areas horizontal edge texture. Experimental results show proposed algorithm outperforms previous...
In this paper an algorithm is presented to extract the valid depth data and correct values of flying pixels by using information confidence image. An adaptive segmentation for measured image executed based on kernel density estimation one-pass connected component labeling. Then a modified structure tensor used detect invalid contained in Finally these are corrected with bi-cubic interpolation method or selectively removed voting operation. And also, erroneous excluded augmented confidence....
In this paper, an effective direction-of-arrival (DOA) estimation method is proposed with a uniform linear array (ULA) when uncorrelated and coherent signals coexist. The direction-of-arrivals (DOAs) of are estimated in two steps. DOAs using conventional subspace firstly, then the information can be eliminated by matrix difference technique, finally decorrelated to estimated. theoretical analysis simulation results show that effective.
Face anti-spoofing plays a crucial role in face recog-nition systems widely used smart devices and security systems. In this paper, we propose multi-stream fusion system based on 3D camera by making full use of information for anti-spoofing. This is composed depth maps Surface Normal Maps (SNM). Detailed discussions about are given. Comparison among different modalities comparison other methods provided through several experiments the public WMCA dataset our self-build Anti-3D dataset. Due...
In this paper, seismic quality factor Q estimation from vertical profile (VSP) data is discussed, by using continuous wavelet transform (CWT). We suppose that source signature a general constant-phase which matches the real one better. Based on CWT of reference and target recording, we derive formula frequency-independent ratio wavelet-domain peak amplitude these two recordings. Wavelet-domain denotes module recording with every fixed scale. The related to dominant frequency standard...
This paper presents a hierarchical and parallel SoC (System on Chip) architecture for vision processor. The computing is divided into 3 task level modules, which are decision, feature reorganization (or pattern generation), extraction. In the proposed architecture, there two separately buses to integrate also new interrupt RISC processor implement synchronization between hardware modules software. human-face detecting tracking application demo has been mapped verified FPGA. Architecture...
Recently, many deep image compression methods have been proposed and achieved remarkable performance. However, these are dedicated to optimizing the performance speed at medium high bitrates, while research on ultra low bitrates is limited. In this work, we propose a enhanced invertible encoding network guided by traditional transformation theory, experiments show that our codec outperforms existing in both reconstruction Specifically, introduce Block Discrete Cosine Transformation model...
Recently, many deep image compression methods have been proposed and achieved remarkable performance. However, these are dedicated to optimizing the performance speed at medium high bitrates, while research on ultra low bitrates is limited. In this work, we propose a enhanced invertible encoding network guided by traditional transformation theory, experiments show that our codec outperforms existing in both reconstruction Specifically, introduce Block Discrete Cosine Transformation model...
This paper introduces a novel benchmark as part of the AIS 2024 Real-Time Image Super-Resolution (RTSR) Challenge, which aims to upscale compressed images from 540p 4K resolution (4x factor) in real-time on commercial GPUs. For this, we use diverse test set containing variety ranging digital art gaming and photography. The are using modern AVIF codec, instead JPEG. All proposed methods improve PSNR fidelity over Lanczos interpolation, process under 10ms. Out 160 participants, 25 teams...
Compressing images at extremely low bitrates (below 0.1 bits per pixel (bpp)) is a significant challenge due to substantial information loss. Existing extreme image compression methods generally suffer from heavy artifacts or low-fidelity reconstructions. To address this problem, we propose novel framework that combines compressive VAEs and pre-trained text-to-image diffusion models in an end-to-end manner. Specifically, introduce latent feature-guided module based on VAEs. This compresses...