Can you find me? By simulating how humans to discover the so-called 'perfectly'-camouflaged object, we present a novel boundary-guided separated attention network (call BSA-Net). Beyond existing camouflaged object detection (COD) wisdom, BSA-Net utilizes two-stream modules highlight separator (or say object's boundary) between an image's background and foreground: reverse stream helps erase interior focus on background, while normal recovers thus pay more foreground; both streams are...

Snow is one of the toughest adverse weather conditions for object detection (OD). Currently, not only there a lack snowy OD datasets to train cutting-edge detectors, but also these detectors have difficulties learning latent information beneficial in snow. To alleviate two above problems, we first establish real-world dataset, named RSOD. Besides, develop an unsupervised training strategy with distinctive activation function, called <inline-formula...

Camouflaged objects share very similar colors but have different semantics with the surroundings. Cognitive scientists observe that both global contour (i.e., boundary) and local pattern texture) of camouflaged are key cues to help humans find them successfully. Inspired by cognitive scientist's observation, we propose a novel boundary-and-texture enhancement network (FindNet) for object detection (COD) from single images. Different most existing COD methods, FindNet embeds information into...

Purpose Existing calibration methods mainly focus on the camera laser-plane of a single laser-line length, which is not convenient and cannot guarantee consistency results when several three-dimensional (3D) scanners are involved. Thus, this study aims to provide unified step for different length requirements laser profile measurement (LPM) systems. Design/methodology/approach 3D LPM process converting physical objects into digital models, wherein critical ensuring system precision. However,...

Small targets are often submerged in cluttered backgrounds of infrared images. Conventional detectors tend to generate false alarms, while CNN-based lose small deep layers. To this end, we propose iSmallNet, a multi-stream densely nested network with label decoupling for object detection. On the one hand, fully exploit shape information targets, decouple original labeled ground-truth (GT) map into an interior and boundary one. The GT map, collaboration two additional maps, tackles unbalanced...

In this study we propose a deformable pattern recognition method with CUDA implementation. order to achieve the proper correspondence between foreground pixels of input and prototype images, pair distance maps are generated from whose pixel values given based on nearest pixel. Then regularization technique computes horizontal vertical displacements these maps. The dissimilarity is measured eight-directional derivative images in leverage characteristic information curvature line segments that...

Optical coherence tomography (OCT) is widely used to diagnose retinal diseases. However, due the limited resolution of OCT imaging systems, quality fundus images displayed not satisfactory, which hinders diagnosis patients by ophthalmologists. This an inevitable problem but few people have given attention it. We attempt solve this through deep learning methods.In paper, we propose a single-image superresolution (SISR) model that based on generative adversarial network (GAN) for restoring...

Abstract Salient target detection in hyperspectral image is a significant task segmentation, tracking, classification and so on. Many existing saliency algorithms for cannot present the boundary of salient well description not enough. A method based on visual attention to detect proposed this paper. In method, frequency‐tuned (FT) model combined with spectral image. FT used get clear border, information made full use improve accuracy detection. Firstly, map generated. Then, measured by...

Chinese characters have distilled the nation's vast wisdom and values, but general public's learning enjoyment of ancient scripts are hampered by fact that fonts from different dynasties highly styles, intricate structures, diverse deformations. To solve difficulty ordinary people identifying characters, an font recognition system is based on improved Inception-ResNet network proposed. ECA-Net integrated into Inception module, PReLU utilized to activate network, Nadam algorithm used enhance...

Aiming at improving noise reduction effect for mechanical vibration signal, a Gaussian mixture model (SGMM) and quantum‐inspired standard deviation (QSD) are proposed applied to the denoising method using thresholding function in wavelet domain. Firstly, SGMM is presented utilized as local distribution approximate coefficients each subband. Then, within Bayesian framework, maximum posteriori (MAP) estimator employed derive with conventional (CSD) which calculated by expectation‐maximization...

Due to the inherent locality of convolutional operations, convolution neural network (CNN) often exhibits limitations in explicitly modeling long-distance dependencies. In this paper, we propose a novel hybrid multi-level graph (HMGN) that combines CNN and capture both local non-local image features at multiple scales. With proposed patch attention module, HMGN can over large receptive field, resulting more accurate segmentation cardiac structures. Experiments on two public datasets show...

Traditional kernel-based object tracking methods are useful for estimating the position of objects, but inadequate scale objects. In this paper, we propose a novel invariant (SIKBOT) algorithm fast scaling objects through image sequences. We exploit set property regions and new method to estimate potential intersection kernel. Regarding robustness, iteratively by means basic analysis. The simultaneously estimated mean shift procedures in parallel. proposed SIKBOT is demonstrated extensive...

Medical image segmentation is always the hot topic of medical analysis. Due to images' complex topological changes, high noise and lower contrast, one-dimensional histogram based classical thresholding methods are helpless. Therefore, 2D histogram-based have been gradually became issue segmentation. Since basic GA maximum fuzzy entropy algorithm has problem premature, this paper uses an improved genetic (IGA) optimize time-consuming question. Through using distance biggest fitness value...

Abstract. Image registration is a fundamental in remote sensing applications such as inter-calibration and image fusion. Compared to other multi sensor problems optical-to-IR, the for SAR optical images has its specials. Firstly, radiometric geometric characteristics are different between images. Secondly, feature extraction methods heavily suffered with speckle Thirdly, structural information more useful than point features corners. In this work, we proposed novel Gaussian Mixture Model...

In order to solve the problems of "ghost" effect and noise interference classical Vibe algorithm in moving object detection, an improved based on visual attention mechanism was put forward. this algorithm, two-dimensional entropy saliency any frame are firstly calculated, by which adaptive background updating factor is derived. Then, model can be adaptively updated according changes between adjacent frames. And also used suppress eliminate ghost rapidly. Comparison experimental results show...

To Evaluate SRM effectively, according to the structure characteristics of solid rocket motor, series motor ICT images were processed with edge detection ,edge thinning, contour tracing, segmentation, and fitting. The raster converted vector which can be recognized by CAD modeling software. Then, images, was modeled software, model studied finite-element analysis. experimental result indicates that quality is good, analysis reflect state SRM.

Detailed 3D reconstruction and photo-realistic relighting of digital humans are essential for various applications. To this end, we propose a novel sparse-view 3d human framework that closely incorporates the occupancy field albedo with an additional visibility field--it not only resolves occlusion ambiguity in multiview feature aggregation, but can also be used to evaluate light attenuation self-shadowed relighting. enhance its training viability efficiency, discretize onto fixed set sample...

Abstract. In order to solve the problem of automatic detection artificial objects in high resolution remote sensing images, a method for areas images based on multi-scale and multi feature fusion is proposed. Firstly, geometric features such as corner, straight line right angle are extracted from original resolution, pseudo corner points, linear orthogonal angles filtered out by self-constraint mutual restraint between them. Then radiation intensity map image with characteristics obtained...