We present a new image editing method, particularly effective for sharpening major edges by increasing the steepness of transition while eliminating manageable degree low-amplitude structures. The seemingly contradictive effect is achieved in an optimization framework making use L 0 gradient minimization, which can globally control how many non-zero gradients are resulted to approximate prominent structure sparsity-control manner. Unlike other edge-preserving smoothing approaches, our method...

A facile method is developed to synthesize intrinsically fluorescent carbon dots by hydrothermal treatment of glucose in the presence monopotassium phosphate. The fluorescence emission thus produced tunable simply adjusting concentration

Complex structures commonly exist in natural images. When an image contains small-scale high-contrast patterns either the background or foreground, saliency detection could be adversely affected, resulting erroneous and non-uniform assignment. The issue forms a fundamental challenge for prior methods. We tackle it from scale point of view propose multi-layer approach to analyze cues. Different varying patch sizes downsizing images, we measure region-based scales. final values are inferred...

Most of the recent successful methods in accurate object detection and localization used some variants R-CNN style two stage Convolutional Neural Networks (CNN) where plausible regions were proposed first then followed by a second for decision refinement. Despite simplicity training efficiency deployment, single have not been as competitive when evaluated benchmarks consider mAP high IoU thresholds. In this paper, we novel end-to-end trainable network to overcome limitation. We achieved...

Recently, visual Transformer (ViT) and its following works abandon the convolution exploit self-attention operation, attaining a comparable or even higher accuracy than CNN. More recently, MLP-mixer abandons both proposing an architecture containing only MLP layers. To achieve cross-patch communications, it devises additional token-mixing besides channel-mixing MLP. It achieves promising results when training on extremely large-scale dataset such as JFT-300M. But cannot outstanding...

In this paper, we propose a novel localization methodology to enhance the accuracy from two aspects, i.e., adapting uncertain noise of microelectromechanical system-based inertial navigation system (MEMS-INS) and accurately predicting INS errors. First, an interacting multiple model (IMM)-based sequential two-stage Kalman filter is proposed fuse information MEMS-INS, global positioning (GPS), in-vehicle sensors. Three bias filters are built with different covariance matrices cover wide range...

Previous joint/guided filters directly transfer the structural information in reference image to target one. In this paper, we first analyze its major drawback -- that is, there may be completely different edges two images. Simply passing all patterns could introduce significant errors. To address issue, propose concept of mutual-structure, which refers is contained both images and thus can safely enhanced by joint filtering, an untraditional objective function efficiently optimized yield...

Many applications have benefited from the images with both high spatial and spectral resolution, such as mineralogy surveillance. However, it is difficult to acquire due limitation of sensor technologies. Recently, super-resolution (SR) techniques been proposed improve or resolution images, e.g., improving hyperspectral (HSIs) color (reconstructing HSIs RGB inputs). none researches attempted together. In this article, these two types are jointly improved using convolutional neural network...

Continuing efforts have been made to explore novel exopolysaccharides (EPSs) for valuable applications. In this research, we report the first time that a non-glucan EPS named EPS-605 can self-assemble form spherical nanosize particles of ∼88 nm in diameter, expanding both range type and structural EPSs into. Characterization shows it is composed mannose, glucose, galactose with several modifications including acylation, phosphorylation, sulfation, carboxylation, highly negative charge....

How to achieve reliable and accurate positioning performance using low-cost sensors is one of the main challenges for land vehicles. This paper proposes a novel fusion strategy vehicles in GPS-denied environments, which enhances simultaneously from sensor methodology levels. It integrates multiple complementary not only incorporating GPS microelectromechanical-based inertial measurement unit, but also "virtual" sensor, i.e., sliding-mode observer (SMO). The SMO first synthesized based on...

In large-scale long-term dynamic environments, high-frequency objects inevitably lead to significant changes in the appearance of scene at same location different times, which is catastrophic for place recognition (PR). Therefore, how eliminate influence achieve robust PR has universal practical value mobile robots and autonomous vehicles. To this end, we suggest a novel semantically consistent LiDAR method based on chained cascade network, called SC_LPR, mainly consists semantic image...

Camera shake during exposure time often results in spatially variant blur effect of the image. The non-uniform is not only caused by camera motion, but also depth variation scene. objects close to sensors are likely appear more blurry than those at a distance such cases. However, recent deblurring methods do explicitly consider factor or assume fronto-parallel scenes with constant for simplicity. While single image challenging problem, fact contain information which can be exploited. We...

Vehicle sideslip and yaw angles are critical for many vehicle safety systems. Although much research has been presented to obtain them individually, simultaneous accurate estimation of them, based on affordable sensors land applications, is seldom reported. This paper proposes a fusion methodology integrating single-frequency double-antenna Global Positioning System (DA-GPS) with other low-cost in-vehicle achieve reliable both angles. The proposed adopts hybrid decentralized filtering...

Single image super resolution (SR), which refers to reconstruct a higher-resolution (HR) from the observed low-resolution (LR) image, has received substantial attention due its tremendous application potentials. Despite breakthroughs of recently proposed SR methods using convolutional neural networks (CNNs), their generated results usually lack preserving structural (high-frequency) details. In this paper, regarding global boundary context and residual as complimentary information for...

We recently have witnessed many ground-breaking results in machine learning and computer vision, generated by using deep convolutional neural networks (CNN). While the success mainly stems from large volume of training data network architectures, vector processing hardware (e.g. GPU) undisputedly plays a vital role modern CNN implementations to support massive computation. Though much attention was paid extent literature understand algorithmic side CNN, little research dedicated...

Photos compress 3D visual data to 2D. However, it is still possible infer depth information even without sophisticated object learning. We propose a solution based on small-scale defocus blur inherent in optical lens and tackle the estimation problem by proposing non-parametric matching scheme for natural images. It incorporates prior with our newly constructed edgelet dataset using non-local scheme, includes semantic order cues physically inference. Several applications are enabled images,...

Color, infrared and flash images captured in different fields can be employed to effectively eliminate noise other visual artifacts. We propose a two-image restoration framework considering input from fields, for example, one noisy color image dark-flashed near-infrared image. The major issue such is handle all structure divergence find commonly usable edges smooth transitions visually plausible reconstruction. introduce novel scale map as competent representation explicitly model...

Automatic speaker naming is the problem of localizing as well identifying each speaking character in a TV/movie/live show video. This challenging mainly attributes to its multimodal nature, namely face cue alone insufficient achieve good performance. Previous approaches this usually process data different modalities individually and merge them using handcrafted heuristics. Such work for simple scenes, but fail high performance speakers with large appearance variations. In paper, we propose...

In this paper, we propose a cost-effective localization solution for land vehicles, which can simultaneously adapt to the uncertain noise of inertial sensors and bridge Global Positioning System (GPS) outages. First, three Unscented Kalman filters (UKFs) with different covariances are introduced into framework Interacting Multiple Model (IMM) algorithm form proposed IMM-based UKF, termed as IMM-UKF. The IMM provide soft switching among UKFs therefore characteristics. Further, two IMM-UKFs...