Camouflaged object detection (COD) aims to segment objects that are visually indistinguishable from their surroundings. The intrinsic low boundary contrast and ambiguity of camouflaged instances pose critical challenges to accurate segmentation. To address the subtle boundaries and ambiguous semantics of camouflaged instances, we propose a edge-guided remapping fusion network (ERNet), a prior-optimized framework that integrates edge cues into the segmentation process, progressively reconstructing object boundaries and refining predictions. Specifically, we introduce an edge detection module (EDM) that employs multi-directional depth-wise convolutions to extract structure-aware edge information. To further enhance feature representation, we design a context-boundary-aware feature aggregation (CFA) module that incorporates spatially refined edge priors into multi-scale contextual integration. Moreover, we propose a learnable edge-mask optimization strategy, which leverages the discrepancy between predicted masks and edge maps as a guidance signal for value-space remapping, enabling more accurate recovery of missing regions near object boundaries. Experiments conducted on four benchmark datasets demonstrate that our ERNet achieves state-of-the-art performance while producing more precise and complete object boundaries. Our code is publicly available at: https://github.com/AaHa123/ERNet.

Camouflaged object detection (COD), identifying objects that visually blended into the surrounding backgrounds, is still a valuable challenge task. While existing methods leverage CNNs or Transformer for feature extraction, they often struggle with limited receptive fields or high computation, and inefficient fusion of multi-domain cues. To this end, we in this paper propose a novel bidirectional position-aware network (BPNet) that synergistically combines spatial-frequency analysis with dynamic contextual learning. Our method expands receptive fields via cascaded wavelet decomposition while preserving computation, and employ adaptive Laplacian filters to amplify subtle discriminative clues within spatial and frequency domains. The framework further progressively refines predictions through ordinary differential equation-inspired learning and bidirectional position map, which dynamically update thresholds through an iterative loop of segmentation, feature learning, and resegmentation. Experiments across four benchmarks demonstrate that BPNet outperforms 15 state-of-the-art methods, validating its effectiveness in balancing global context capture and fine-grained detail preservation for complex COD scenarios. Our code is publicly available at: https://github.com/AaHa123/BPNet.

Addressing the critical challenges of behavior feature loss, privacy protection, and cross-modal analysis in low-resolution teaching videos, this work proposes a teaching video super-resolution reconstruction framework based on cross-modal feature Mamba fusion. This approach integrates G-buffer information from computer graphics and video temporal features, creating a dual-branch architecture. The encoding end employs a dynamic feature selection mechanism to enhance spatial structure modeling, while the decoding end uses the Mamba network and SS2D module for feature extraction and spatio-temporal feature aggregation to reduce jitter between frames. Additionally, sub-pixel convolution is used to optimize the reconstruction of high-frequency details. Experimental results show that the TVSR framework reduces 51% of the parameter count and 15% of the latency compared to SOTA methods, while significantly improving the clarity of key teaching elements such as whiteboard writing and teaching tool operations. This research aims to expand super-resolution technology into the field of educational digitalization, providing technical support for the three-tier cognitive system, which construction of the physical classroom, digital mirror and teaching intelligence.

Abstract Lung segmentation, a prerequisite step of lung disease detection in computer‐aided diagnosis system, is a challenging task because of noises, complex structures, as well as large individual differences of lung CT scans. Here, an automatic algorithm for segmenting lungs from thoracic CT images accurately is presented. This scheme consists of three principal steps: image preprocessing, lung extracting and contour correcting. To cope with inhomogeneous intensities of CT images, a novel preprocessing approach based on empirical mode decomposition and bilateral filter is proposed, which has abilities of denoising, smoothing and edge keeping. Lung region is then extracted with a novel gray correlation‐based clustering approach. A new lung contour correction technology is finally employed to repair the concave regions caused by pulmonary nodules, vessels and so on. Experimental results show that the preprocessing approach outperforms other methods on image denoising and smoothing. Meanwhile, the lung segmentation algorithm is tested on a group of lung CT images affected with interstitial lung diseases and achieves a high segmentation accuracy. Compared with several existing lung segmentation methods, this algorithm exhibits a better performance on lung segmentation.

Accurately segmenting lungs from CT images is a fundamental step for quantitative analysis of lung diseases. However, it is still a challenging task because of some interferential factors, such as juxta‐pleural nodules, pulmonary inflammation, as well as individual anatomical varieties. In this study, with the combination of a superpixel approach and a hybrid model composed of convolutional neural network and random forest (CNN‐RF), the authors propose a novel algorithm to segment lungs from CT images in an automatic and accurate fashion. The authors' lung segmentation covers three main stages: image preprocessing, lung segmenting and segmentation refining. A lung CT image denoised with a fractional‐order grey similarity approach is first segmented to a set of superpixels, and the CNN‐RF model is then employed to classify the superpixels and identify lungs from the CT image. The segmentation result is further refined by separating the left and right lungs, eliminating trachea, and correcting lung contours. Experiments show that their algorithm can generate more accurate lung segmentation results with 94.98% Jaccard's index and 97.99% Dice similarity coefficient, compared with ground truths, and it achieved better results compared with several feature‐based machine learning techniques and current methods on lung segmentation.

To achieve an automatic and accurate segmentation of lungs and improve the clinical efficiency of computer‐aided diagnosis, the authors present a lung segmentation algorithm based on the random forest method and a multi‐scale edge detection technique. The algorithm carries a first step of lung region extraction and a second step of lung nodule segmentation. By combining texture information, the improved superpixel generation method can better deal with initial segmentation on lung computed tomography images with inhomogeneous intensity. Then, the lung region is further extracted by using the random forest classifier on the superpixel features, and the lung contours are corrected with a proposed circle tracing technique. Finally, the segmentation is further refined by employing a multi‐scale edge detection technique, which enables their method to detect suspicious nodules with various intensities and sizes adaptively. The effectiveness of the proposed approach is demonstrated on a group of datasets by comparing with the corresponding ground truths as well as the classical algorithms. Experimental results show that the proposed method has a higher precision than the compared algorithms in a fully automatic fashion.

Bas-relief is characterized by its unique presentation of intrinsic shape properties and/or detailed appearance using materials raised up in different degrees above a background. However, many bas-relief modeling methods could not manipulate scene details well. We propose a simple and effective solution for two kinds of bas-relief modeling (i.e., structure-preserving and detail-preserving) which is different from the prior tone mapping alike methods. Our idea originates from an observation on typical 3D models, which are decomposed into a piecewise smooth base layer and a detail layer in normal field. Proper manipulation of the two layers contributes to both structure-preserving and detail-preserving bas-relief modeling. We solve the modeling problem in a discrete geometry processing setup that uses normal-based mesh processing as a theoretical foundation. Specifically, using the two-step mesh smoothing mechanism as a bridge, we transfer the bas-relief modeling problem into a discrete space, and solve it in a least-squares manner. Experiments and comparisons to other methods show that (i) geometry details are better preserved in the scenario with high compression ratios, and (ii) structures are clearly preserved without shape distortion and interference from details.

In this paper, we propose a multiple-view matrix based 3D reconstruction algorithm for generating a 3D point cloud model for a scene or an object from several sequence images. The algorithm first extracts a group of SIFT (Scale Invariant Feature Transform) feature points from each image, and divides the points into different groups according to the matching degrees among the points. Secondly, a set of 3D point clouds are reconstructed from the feature points with a calculated a multiple-view matrix. Then, a complete result is generated by merging the point clouds with an incremental algorithm and the estimated camera parameters. Furthermore, our result is optimized by employing a BA (Bundle Adjustment) method. Owing to the introduction of the multiple-view matrix and the group-based SIFT matching, our algorithm has the ability to accurately reconstruct a 3D point cloud model only with several images. The performance of our algorithm is evaluated on a group of benchmark datasets, and is compared to two state-of-the-art methods.

Multi-body collision detection is a key and impor-tant technology in societies of computer graphics, system simu-lation, virtual reality, etc, and has been widely used in various applications. To deal with the collision problems in large scale multi-body simulations robustly and efficiently, we in this paper proposed a robust and efficient algorithm of continuous multi-body collision detection based on the kinetic "Sweep and Prune" (SaP) technique and the event-driven mechanism. Our algorithm first culls redundant detection calculations among very large numbers of moving bodies, and then automatically generates events to predict these collisions, probably taken place in coming time, of the object pairs. All these events are been pushed into a priority queue, which is used to drive our algorithm to run. By introducing a new hybrid bounding box hierarchy in the event processing process, our algorithm can detect positions where the object pairs collide. We discovered the event blocking problem potentially occurred during event processing, and further proposed several methods to alarm or relieve the system from the event blocking state. Experimental results show that our algorithm has good stability and strong robustness, and it can improve the speed and accuracy of the multi-body collision detection effectively.

How to improve strength-to-weight ratio of printed models is an important topic in 3D printing. We in this paper propose a novel structure modeling method based on the implicit function technique and the finite element method (FEM). Our method first obtains a set of sampled points in a given surface mesh model by using a probability-based strategy, and generates an adaptive tetrahedral mesh from the points. FEM is then used to analyze the stress of the tetrahedral mesh and a stress map of the input model is created. The method finally builds a result model composed of a shell and an interior single-gyroid lattice. The lattice is defined by a piecewise 3D implicit function, and has several special structural properties just like the structure of the light but strong butterfly wing. The lattice together with the shell forms a natural 3D structure for 3D printing. Local thickness of lattice rods in the structure adaptively changes with stress distribution for withstanding external loads. Experimental results show that our method can deal with various surface mesh models in rapid way, and the resulted models for 3D printing have high strength-to-weight ratios.

Implicit curves play an essential role in the societies of medicine, meteorology, geology, geo-physics, visualization and so on. In this paper, we propose an algorithm to visualize implicit curves and reproduce their sharp features in 2D plane. To access the subdivision cells of a user-defined 2D domain, our algorithm first creates a quadtree by using a top-down and adaptive quad-tree construction technique. In each cell, the method locates exact one feature point of the numerical field defined by the implicit function defining an implicit curve. A discrete optimization technique is employed to calculate the feature points. A dual mesh is subsequently constructed for the quadtree by taking the feature points as its vertices. Our algorithm approximates local part of the implicit curve in each cell of the dual mesh with a modified version of the marching squares method. Collecting all the approximations in the cells, our method finally reproduces the implicit curve with sharp features. Experiments show that our method can efficiently extract the sharp features of implicit curves, and it can work with various implicit curves with or without sharp features robustly.

Point cloud classification plays a critical role in many applications of airborne light detection and ranging (LiDAR) data. In this paper, we present a deep feature-based method for accurately classifying multiple ground objects from airborne LiDAR point clouds. With several selected attributes of LiDAR point clouds, our method first creates a group of multi-scale contextual images for each point in the data using interpolation. Taking the contextual images as inputs, a multi-scale convolutional neural network (MCNN) is then designed and trained to learn the deep features of LiDAR points across various scales. A softmax regression classifier (SRC) is finally employed to generate classification results of the data with a combination of the deep features learned from various scales. Compared with most of traditional classification methods, which often require users to manually define a group of complex discriminant rules or extract a set of classification features, the proposed method has the ability to automatically learn the deep features and generate more accurate classification results. The performance of our method is evaluated qualitatively and quantitatively using the International Society for Photogrammetry and Remote Sensing benchmark dataset, and the experimental results indicate that our method can effectively distinguish eight types of ground objects, including low vegetation, impervious surface, car, fence/hedge, roof, facade, shrub and tree, and achieves a higher accuracy than other existing methods.

We present a rapid and effective point simplification algorithm for surface reconstruction which can represent different levels-of-detail. The core of this algorithm is to generate an approximately minimal set of adaptive balls covering the whole surface by defining and minimizing local quadric error functions. First, the feature points are extracted by simple thresholding curvatures, Second, for the non-feature points, they are covered by distinct balls. The size of each ball varies and reflects how curved the local surface is. Once the size of radius is fixed, the points in each ball will be substituted by an optimized point. Thus, the simplified surface consists of extracted feature points and optimized points. we can employ this algorithm to produce coarse-to-fine models by controlling a general error level, and name it as ESimp for short. Worthy of note, the error level of each ball may be adaptively adjusted according to the local curvature and density of the center of this ball which can avoid holes generation. Finally, the simplified points are triangulated by Cocone algorithm. This algorithm has been applied to a set of large scanned models. Experimental results demonstrate that it can generate high-quality surface approximation with feature preservation.

Various Partial Differential Equations (PDE) have been used in computer graphics for approximating surfaces of geometric shapes by finding solutions to PDEs subject to suitable boundary conditions. The PDE boundary conditions are defined as 3D curves on the surface of the shapes. We propose how to automatically derive these curves as boundaries of curved patches on the surface of the original polygon mesh. The analytic solution to the PDE used throughout this work is fully determined by finding a set of coefficients associated with parametric functions according to the particular set of boundary conditions. The PDE coefficients require an order of magnitude smaller space compared to the original polygon data and can be interactively rendered with different level of detail. It allows for an efficient exchange of the PDE shapes in 3D Cyber worlds and their web visualization. In this paper we analyze and formulate the requirements for extracting suitable boundary conditions, describe the algorithm for the automatic deriving of the boundary curves, and present its implementation as a part of the function-based extension of VRML and X3D.

In this paper, based on quadric error metrics, we present an hybrid approach to optimize triangulation created from implicit surface with sharp features. The approach first uses a resampling process to update vertices positions of initial triangulation. A dual mesh of the triangulation is at the same time constructed to restrict the updated positions and to project the new positions onto the implicit surface. Each vertex position of the updated triangulation is then optimized by minimizing the squared distances from the vertex to tangent planes of the implicit surface at the corresponding vertex of the modified dual mesh. The optimization process combines with a curvature-dependent adaptive mesh subdivision. Our method uses an error metrics to measure deviation between the vertices in final triangulation and the implicit surface.

In this paper, a mathematical model of strategic investment system is first built based on the feature classification method and transition probability of changes describing investment intents of investors. Graphic simulation tool for numerical calculation is then employed to simulate and solve the model. Some calculating results are obtained, which describes long-term behaviors of dynamic evolution of the system with respect to various initial conditions and structure parameters. Related analysis about the simulations can give some theoretical suggestions for investment benefit prediction and macroscopical management for economy. The modelling and simulating method in the paper can also be used as references for studying space structure change and time-based vibration process, of other social systems.

In this paper, we present a novel method to render colorful 3D fractals based on so-called color-control sphere (CSS) and our improved escape-time algorithm. In the method, each point, called parameter point, in parameter domain is first judged whether it is an escape-point or not by an iteration loop, under the governing of a discrete dynamic system. Once the iteration loop is exited, an orbiting point corresponding to the parameter point is then mapped to a mapped-point in CCS and the distance from the mapped-point to the center of CCS is evaluated. Subsequently, a color for rendering the parameter point can be obtained by employing a predefined color-control function in CCS with respect to the distance. Finally, all surface points, which are taken as a set of initial points of the discrete dynamic system, of arbitrary shape embedded in the parameter domain are rendered with their corresponding colors. As a result, a pseudo-3D fractal can be archived on the 3D surface.