A5101457342- Advanced Image and Video Retrieval Techniques
- Advanced Vision and Imaging
- Image Retrieval and Classification Techniques
- Face and Expression Recognition
- Face recognition and analysis
- Video Analysis and Summarization
- Video Surveillance and Tracking Methods
- Image and Signal Denoising Methods
- Advanced Image Processing Techniques
- Human Pose and Action Recognition
- Speech and Audio Processing
- Music and Audio Processing
- Image Processing Techniques and Applications
- Advanced Neural Network Applications
- Robotics and Sensor-Based Localization
- Sparse and Compressive Sensing Techniques
- Optical measurement and interference techniques
- Domain Adaptation and Few-Shot Learning
- Medical Image Segmentation Techniques
- Hand Gesture Recognition Systems
- Emotion and Mood Recognition
- Anomaly Detection Techniques and Applications
- Neural Networks and Applications
- Advanced Data Compression Techniques
- Remote-Sensing Image Classification
University of Illinois Urbana-Champaign
2013-2023
Central South University
2023
Jet Propulsion Laboratory
2004-2022
International University of the Caribbean
2018-2021
Nature Inspires Creativity Engineers Lab
2010-2020
York University
2019-2020
University of Michigan–Ann Arbor
2020
Seoul National University
2019
Kapiolani Medical Center for Women and Children
2019
Nanjing University of Science and Technology
2016
This paper presents a new approach to single-image super-resolution, based on sparse signal representation. Research image statistics suggests that patches can be well-represented as linear combination of elements from an appropriately chosen over-complete dictionary. Inspired by this observation, we seek representation for each patch the low-resolution input, and then use coefficients generate high-resolution output. Theoretical results compressed sensing suggest under mild conditions,...
Two point sets {pi} and {p'i}; i = 1, 2,..., N are related by p'i Rpi + T Ni, where R is a rotation matrix, translation vector, Ni noise vector. Given {p'i}, we present an algorithm for finding the least-squares solution of T, which based on singular value decomposition (SVD) 3 × matrix. This new compared to two earlier algorithms with respect computer time requirements.
The traditional SPM approach based on bag-of-features (BoF) requires nonlinear classifiers to achieve good image classification performance. This paper presents a simple but effective coding scheme called Locality-constrained Linear Coding (LLC) in place of the VQ SPM. LLC utilizes locality constraints project each descriptor into its local-coordinate system, and projected coordinates are integrated by max pooling generate final representation. With linear classifier, proposed performs...
Recently SVMs using spatial pyramid matching (SPM) kernel have been highly successful in image classification. Despite its popularity, these nonlinear a complexity O(n <sup xmlns:mml="http://www.w3.org/1998/Math/MathML" xmlns:xlink="http://www.w3.org/1999/xlink">2</sup> ∼ n xmlns:xlink="http://www.w3.org/1999/xlink">3</sup> ) training and O(n) testing, where is the size, implying that it nontrivial to scaleup algorithms handlemore than thousands of images. In this paper we develop an...
Matrix factorization techniques have been frequently applied in information retrieval, computer vision, and pattern recognition. Among them, Nonnegative Factorization (NMF) has received considerable attention due to its psychological physiological interpretation of naturally occurring data whose representation may be parts based the human brain. On other hand, from geometric perspective, is usually sampled a low-dimensional manifold embedded high-dimensional ambient space. One then hopes...
Recent deep learning based approaches have shown promising results for the challenging task of inpainting large missing regions in an image. These methods can generate visually plausible image structures and textures, but often create distorted or blurry textures inconsistent with surrounding areas. This is mainly due to ineffectiveness convolutional neural networks explicitly borrowing copying information from distant spatial locations. On other hand, traditional texture patch synthesis are...
Techniques from sparse signal representation are beginning to see significant impact in computer vision, often on nontraditional applications where the goal is not just obtain a compact high-fidelity of observed signal, but also extract semantic information. The choice dictionary plays key role bridging this gap: unconventional dictionaries consisting of, or learned from, training samples themselves provide obtaining state-of-the-art results and attaching meaning representations....
Content-based image retrieval (CBIR) has become one of the most active research areas in past few years. Many visual feature representations have been explored and many systems built. While these efforts establish basis CBIR, usefulness proposed approaches is limited. Specifically, relatively ignored two distinct characteristics CBIR systems: (1) gap between high-level concepts low-level features, (2) subjectivity human perception content. This paper proposes a relevance feedback based...
We present a generative image inpainting system to complete images with free-form mask and guidance. The is based on gated convolutions learned from millions of without additional labelling efforts. proposed convolution solves the issue vanilla that treats all input pixels as valid ones, generalizes partial by providing learnable dynamic feature selection mechanism for each channel at spatial location across layers. Moreover, masks may appear anywhere in any shape, global local GANs designed...
This paper addresses the problem of generating a super-resolution (SR) image from single low-resolution input image. We approach this perspective compressed sensing. The is viewed as downsampled version high-resolution image, whose patches are assumed to have sparse representation with respect an over-complete dictionary prototype signal-atoms. principle sensing ensures that under mild conditions, can be correctly recovered signal. will demonstrate effectiveness sparsity prior for...
We present a fast algorithm for two-dimensional median filtering. It is based on storing and updating the gray level histogram of picture elements in window. The much faster than conventional sorting methods. For window size m × n, computer time required 0(n).
This paper reviews the first challenge on single image super-resolution (restoration of rich details in an low resolution image) with focus proposed solutions and results. A new DIVerse 2K dataset (DIV2K) was employed. The had 6 competitions divided into 2 tracks 3 magnification factors each. Track 1 employed standard bicubic downscaling setup, while unknown operators (blur kernel decimation) but learnable through high res train images. Each competition ∽100 registered participants 20 teams...
Two main results are established in this paper. First, we show that seven point correspondences sufficient to uniquely determine from two perspective views the three-dimensional motion parameters (within a scale factor for translations) of rigid object with curved surfaces. The points should not be traversed by planes one plane containing origin, nor cone origin. Second, set ``essential parameters'' introduced which up translations, and can estimated solving linear equations derived eight...
In this paper, we propose a novel coupled dictionary training method for single image super-resolution based on patchwise sparse recovery, where the learned couple dictionaries relate low- and high-resolution patch spaces via representation. The learning process enforces that representation of low-resolution in terms can well reconstruct its underlying with highresolution space. We model problem as bilevel optimization problem, includes an 1-norm minimization constraints. Implicit...
Technology advances in the areas of image processing (IP) and information retrieval (IR) have evolved separately for a long time. However, successful content-based systems require integration two. There is an urgent need to develop mechanisms link model text model, such that well established techniques can be utilized. Approaches converting feature vectors (IF domain) weighted-term (IR are proposed this paper. Furthermore, relevance feedback technique from IR domain used demonstrate...
Bottom-up human pose estimation methods have difficulties in predicting the correct for small persons due to challenges scale variation. In this paper, we present HigherHRNet: a novel bottom-up method learning scale-aware representations using high-resolution feature pyramids. Equipped with multi-resolution supervision training and aggregation inference, proposed approach is able solve variation challenge multi-person localize keypoints more precisely, especially person. The pyramid...
Deep learning techniques have been successfully applied in many areas of computer vision, including low-level image restoration problems. For super-resolution, several models based on deep neural networks recently proposed and attained superior performance that overshadows all previous handcrafted models. The question then arises whether large-capacity data-driven become the dominant solution to ill-posed super-resolution problem. In this paper, we argue domain expertise represented by...