A5100374071- Image and Signal Denoising Methods
- Advanced Image Processing Techniques
- Generative Adversarial Networks and Image Synthesis
- Advanced Image Fusion Techniques
- Vehicle License Plate Recognition
- Advanced Vision and Imaging
- Computer Graphics and Visualization Techniques
- Image Enhancement Techniques
- Image Retrieval and Classification Techniques
- Medical Image Segmentation Techniques
- Infrared Target Detection Methodologies
- Digital Image Processing Techniques
- Industrial Vision Systems and Defect Detection
- Video Analysis and Summarization
- Image and Object Detection Techniques
- Advanced Image and Video Retrieval Techniques
- Advanced Optical Imaging Technologies
- Multimodal Machine Learning Applications
- Image and Video Stabilization
- Biomedical Text Mining and Ontologies
- Advanced Neural Network Applications
- Ultrasound Imaging and Elastography
- Subtitles and Audiovisual Media
- Interactive and Immersive Displays
- Optical measurement and interference techniques
Huazhong University of Science and Technology
2024
Institute of Physics
2023
Tianjin University
2023
Microsoft (United States)
2023
Chongqing University of Technology
2023
Huawei Technologies (Sweden)
2022
OmniVision Technologies (United States)
2021
Lanzhou Jiaotong University
2020
Technische Universität Ilmenau
2017-2019
University of Dayton
2014-2018
Recently, there has been a growing interest in 3D CNN-based stereo matching methods due to their remarkable accuracy. However, the high complexity of convolution makes it challenging strike balance between accuracy and speed. Notably, explicit volumes contain considerable redundancy. In this study, we delve into more compact 2D implicit network eliminate redundancy boost real-time performance. simply replacing networks with causes issues that can lead performance degradation, including loss...
The SoccerNet 2022 challenges were the second annual video understanding organized by team. In 2022, composed of 6 vision-based tasks: (1) action spotting, focusing on retrieving timestamps in long untrimmed videos, (2) replay grounding, live moment an shown a replay, (3) pitch localization, detecting line and goal part elements, (4) camera calibration, dedicated to intrinsic extrinsic parameters, (5) player re-identification, same players across multiple views, (6) object tracking, tracking...
To alleviate the vocabulary problem, this paper investigates role of user term feedback in interactive text-based image retrieval. Term refers to from a on specific terms regarding their relevance target image. Previous studies have indicated effectiveness text retrieval [14]. However, has not shown be effective our experiments Our results indicate that, although positive effect by allowing users identify more relevant terms, it also strong negative providing opportunities for specify...
We propose corrupted reference image quality assessment (CRIQA), a novel foundation for reasoning about and denoising problems jointly. In order to assess the visual of processed relative an ideal (not provided), we predict full-reference (FRIQA) scores denoised images without having direct access image, but with help observed instead. Our simulation studies verify that CRIQA indeed agree corresponding FRIQA scores, human subject confirm are more consistent perceived than NRIQA scores....
Many important data mining problems can be modeled as learning a (bidirectional) multidimensional mapping between two domains. Based on the generative adversarial networks (GANs), particularly conditional ones, cross-domain joint distribution matching is an increasingly popular kind of methods addressing such problems. Though significant advances have been achieved, there are still main disadvantages existing models, i.e., requirement large amount paired training samples and notorious...
Variational autoencoders (VAEs) have witnessed great success in performing the compression of image datasets. This success, made possible by bits-back coding framework, has produced competitive performance across many benchmarks. However, despite this, VAE architectures are currently limited a combination practicalities and ratios. That is, not only do state-of the-art methods, such as normalizing flows, often demonstrate out-performance, but initial bits required makes single parallel...
Text queries are natural and intuitive for users to describe their information needs. However, text-based image retrieval faces many challenges. Traditional text techniques on descriptions have not been very successful. This is mainly due the inconsistent textual discrepancies between user terms in descriptions. To investigate strategies alleviate this vocabulary problem, article examines role of term feedback targeted search that based retrieval. Term refers from a specific regarding...
At present, the traffic engineering and automation have developed, vehicle license plate recognition technology need get a corresponding improvement also.In case of identifying car picture, principle automatic is illustrated in this paper, processing described detail which includes preprocessing, edge extraction, location, character segmentation, recognition.The program implementing edited by Matlab.The example result shows that method feasible, it can be put into practice.
This study presents a hierarchical palmprint feature extraction and recognition approach based on multi‐wavelet complex network (CN) since they can effectively decrease redundant information enhance key points of main lines wrinkles. The is first pre‐filtered decomposed once using multi‐wavelet. Three components (LL 1,2,3 ) corresponding to the pre‐filter except for diagonal component are extracted as elementary features. Second, binary images (BLL obtained by average window method different...
Video dubbing aims to translate the original speech in a film or television program into target language, which can be achieved with cascaded system consisting of recognition, machine translation and synthesis. To ensure translated well aligned corresponding video, length/duration should as close possible that speech, requires strict length control. Previous works usually control number words characters generated by model similar source sentence, without considering isochronicity duration...
At present, the traffic engineering and automation have developed, vehicle license plate recognition technology need get a corresponding improvement also. In case of identifying car picture, principle automatic is illustrated in this paper, processing described detail which includes pre-processing, edge extraction, location, character segmentation, recognition. The program implementing edited by Matlab. example result shows that method feasible, it can be put into practice.
The distribution of real camera sensor is well approximated by Poisson, and the estimation light intensity signal from Poisson count data plays a prominent role in digital imaging. It highly desirable for imaging devices to carry ability assess performance image restoration. Drawing on new category quality assessment called corrupted reference (CR-QA), we develop computational technique predicting score popular structural similarity index (SSIM) without having direct access ideal image. We...
Generative Adversarial Networks (GANs) have made great success in synthesizing high-quality images. However, how to steer the generation process of a well-trained GAN model and customize output image is much less explored. It has been recently found that modulating input latent code used GANs can reasonably alter some variation factors image, but such manipulation usually presents change entire as whole. In this work, we propose an effective approach, termed LoGAN, support local editing...
In the low-photon imaging regime, noise in image sensors is dominated by shot noise, best modeled statistically as Poisson distribution. this work, we show that likelihood function very well matched with Bayesian estimation of "difference log contrast pixel intensities." More specifically, our work rooted statistical compositional data analysis, whereby reinterpret Aitchison geometry a multi-resolution analysis log-pixel domain. We demonstrate difference-log-contrast has wavelet-like...
Aiming at the problem that traditional prior information‐based defogging algorithm fails in some special scenarios, an end‐to‐end convolutional network based on attention mechanism is proposed. The consists of two modules: parameter estimation and image restoration. First, multi‐scale convolution used to extract feature information. Residual skip connection methods are improve utilisation rate shallow Secondly, channel domain add weight input from previous select useful Finally, atmospheric...
Since detection result is not good enough when detecting real image using line search mechanism corner algorithm, a feature selection criterion based on priority and adaptive non-maximal suppression (ANMS) proposed in this paper to control the number density of features image. Experimental results show that algorithm can detect more reasonable, be used mosaics well.
Surface reconstruction has traditionally relied on the Multi-View Stereo (MVS)-based pipeline, which often suffers from noisy and incomplete geometry. This is due to that although MVS been proven be an effective way recover geometry of scenes, especially for locally detailed areas with rich textures, it struggles deal low texture large variations illumination where photometric consistency unreliable. Recently, Neural Implicit Reconstruction (NISR) combines surface rendering volume techniques...
Invisible watermarking is essential for safeguarding digital content, enabling copyright protection and content authentication. However, existing methods fall short in robustness against regeneration attacks. In this paper, we propose a novel method called FreqMark that involves unconstrained optimization of the image latent frequency space obtained after VAE encoding. Specifically, embeds watermark by optimizing images then extracts through pre-trained encoder. This allows flexible...
The examination of the Computer Graphics is basically computer to investigate drawing ability in universities recent years.Based on many years teaching practice and according transformation trend intelligent paper marking, image processing technology adopted, key information extracted, similarity calculation program compiled, CAD automatic marking function implemented by contrast students' plots with standard answer.Through examples, grading results are consistent artificial ideally.The...
In conventional binaural rendering a pair of head-related impulse responses (HRIR), measured from source direction to left and right ears, is convolved with signal create the impression virtual 3D sound when played on headphones. It well known that using HRIRs in real room, which includes natural reverberant decay, increases externalization realism simulation. However, HRIR filter length even small room can be many thousands taps leading computational complexity issues world implementations....
How to preserve the spectral information when enhancing spatial details is a key issue of remote sensing image fusion. The component substitution (CS)-based fusion methods can effectively enhance while suffering distortion, and multiresolution analysis (MRA)-based have advantages in preserving but are not satisfactory terms details. This paper proposes hybrid method integrate CS- MRA-based approaches. intensity first obtained from an original multispectral (MS) by hyperspherical color space...
For the pores edge detecting of Activated Carbon Fibers (ACF) material images, traditional approaches are difficult to obtain complete information. Snake algorithm is a reasonable approach for detection. An improved initial contour model proposed in this paper. A rectangle first located surround be detected instead drawing series points as contour. Then, we map these on surrounded according certain rule constitute After mapping strategy, used iterate Experiments show that information...
In this paper, we propose a method of data processing for swept volumetric display and its experimental result. The proposed consists four main parts: acquiring, pre-processing, transmitting, post-processing. Different acquisition techniques is adopted various types. Data pre-processing mainly include normalization, Coordinate transformation, reduction & uniform, splitting. post-processing part realizes three functions: receiving broadcasting, saving, transmitting to driver circuit. Compared...
This paper proposes a new approach of automated toll gate passing an driving vehicle. enables the vehicle to select optimal and automatically pass by using object detection, 3D environment construction, virtual line generation, path planning motion control. After designing concept approach, some demonstrations are conducted prove it. data-based scenario shows that proposed can not only perceive well for this purpose but also plan appropriate trajectories when encountering complex scene near plazas.