Search

Recent work has shown that Binarized Neural Networks (BNNs) are able to greatly reduce computational costs and memory footprints, facilitating model deployment on resource-constrained devices. However, in comparison to their full-precision counterparts, BNNs suffer from severe accuracy degradation. Research aiming to reduce this accuracy gap has thus far largely focused on specific network architectures with few or no 1 × 1 convolutional layers, for which standard binarization methods do not work well. Because 1 × 1 convolutions are common in the design of modern architectures (e.g. GoogleNet, ResNet, DenseNet), it is crucial to develop a method to binarize them effectively for BNNs to be more widely adopted. In this work, we propose an “Elastic-Link” (EL) module to enrich information flow within a BNN by adaptively adding real-valued input features to the subsequent convolutional output features. The proposed EL module is easily implemented and can be used in conjunction with other methods for BNNs. We demonstrate that adding EL to BNNs produces a significant improvement on the challenging large-scale ImageNet dataset. For example, we raise the top-1 accuracy of binarized ResNet26 from 57.9% to 64.0%. EL also aids con-vergence in the training of binarized MobileNet, for which a top-1 accuracy of 56.4% is achieved. Finally, with the integration of ReActNet, it yields a new state-of-the-art result of 71.9% top-1 accuracy.

Background The imperfect material effect is one of the most important themes to obtain photo-realistic results in rendering. Textile material rendering has always been a key area in the field of computer graphics. So far, a great deal of effort has been invested in its unique appearance and physicsbased simulation. The appearance of the dyeing effect commonly found in textiles has received little attention. This paper introduces techniques for simulation of staining effects on textiles. Pulling, wearing, squeezing, tearing, and breaking effects are more common imperfect effects of fabrics, these external forces will cause changes in the fabric structure, thus affecting the diffusion effect of stains. Based on the microstructure of yarn, we handle the effect of the stain on the imperfect textile surface. Our simulation results can achieve a photo-realistic effect.

Background The interaction of gas and liquid can produce many interesting phenomena, such as bubbles rising from the bottom of the liquid. The simulation of two-phase fluids is a challenging topic in computer graphics. To animate the interaction of a gas and liquid, MultiFLIP samples the two types of particles, and a Euler grid is used to track the interface of the liquid and gas. However, MultiFLIP uses the fluid implicit particle (FLIP) method to interpolate the velocities of particles into the Euler grid, which suffer from additional noise and instability. Methods To solve the problem caused by fluid implicit particles (FLIP), we present a novel velocity transport technique for two individual particles based on the affine particle-in-cell (APIC) method. First, we design a weighed coupling method for interpolating the velocities of liquid and gas particles to the Euler grid such that we can apply the APIC method to the simulation of a two-phase fluid. Second, we introduce a narrowband method to our system because MultiFLIP is a time-consuming approach owing to the large number of particles. Results Experiments show that our method is well integrated with the APIC method and provides a visually credible two-phase fluid animation. Conclusions The proposed method can successfully handle the simulation of a twophase fluid.

Diabetes mellitus and chronic kidney disease are two severe chronic diseases in the world, affecting the quality of a patient’s life. However, detecting these two diseases often applies professional medical techniques such as a Fasting Plasma Glucose test and estimating the glomerular filtration rate (eGFR) measurement, which usually requires a blood test. Given the various inconveniences and risks in existing conventional diagnostic approaches, noninvasive healthcare systems based on intelligent electronic detection/prevention are preferred. To achieve this goal, we propose a progressively trainable network, i.e., dual stack network (DsNet), to distinguish patients with chronic kidney disease, diabetes mellitus from healthy people simultaneously through analyzing the facial images of candidates. The first stack subnetwork extracts high-level representative features from the facial images effectively. While the second stack subnetwork can further analyze the extracted high-level features from the first stack subnetwork, before classifying the two diseases from healthy individuals simultaneously. Extensive experiments on a dataset with 229 healthy samples, 236 diabetes, and 200 chronic kidney disease patients show that our proposed method generated the F1-score of 95.33%, 98.17%, and 94.67% for detecting chronic kidney disease, diabetes, and healthy samples respectively. Our proposed DsNet achieves significant improvements compared with other traditional noninvasive detection approaches.

Most motion-blurred images captured in the real world have spatially-varying point-spread functions, and some are caused by different positions and depth values, which cannot be handled by most state-of-the-art deblurring methods based on deconvolution. To overcome this problem, we propose a depth-aware motion blur model that treats a blurred image as an integration of a sequence of clear images. To restore the clear latent image, we extend the Richardson-Lucy method to incorporate our blur model with a given depth image. The empty holes in the depth image, caused by occlusion or device limitations, are fixed by PatchMatch-based depth filling. We regard the depth image as a Markov random field and select candidate labels by using belief propagation to set and smooth depth values for empty areas. Deblurring and depth filling are performed iteratively to refine the results. Our method can also be applied to real-world images with the assistance of motion estimation. The deblurring process is shown to be convergent; moreover, the number of iterations and the level of noise amplification are acceptable. The experimental results show that our method can not only handle depth-variant motion blur but also refine depth images.

The epicardial fat plays a key role in the development of many cardiovascular diseases. It is necessary and useful to precisely segment this fat from CT scans in clinical studies. However, it is not feasible to manually segment this fat in clinical practice, as the workload and cost for technicians or physicians is quite high. In this work, we propose a novel method for automatic segmentation and quantification of epicardial fat from CT scans accurately. In detail, dual U-Nets with the morphological processing layer is used for this goal. The first network is based on the U-Net framework to detect the pericardium, before segmenting its inside region. A morphological layer is concatenated as the following layer of the first network, to refine and obtain the ideal inside region of the pericardium. While the second network is also applied using U-Net as its backbone to find and segment the epicardial fat of the processed inside region from the pericardium using the first network. Our proposed method obtains the highest mean Dice similarity (91.19%), correlation coefficient (0.9304) compared to other state-of-art methods on a cardiac CT dataset with 20 patients. The results indicate our proposed method is effective for quantifying epicardial fat automatically.

In this work, we present a supervised object segmentation algorithm for unconstrained video. Instead of arbitrarily picking a few frames for manual labeling, as in many existing supervised methods, the proposed method selects frames in a more reasonable manner, called supervision optimization. For this, we formulate a principled objective function by inferring the propagation error from appearance and motion clues. After this, we construct a multilevel segmentation model, which consists of low-level and high-level features. On the low level, image pixels are used for a more accurate estimation of motion and segmentation. On the high level, image segments are considered for a more semantic classification of the foreground and background. By integrating these in one segmentation graph, the result can be further improved by leveraging the knowledge from both levels. In experiments, the proposed approach is evaluated by different measures, and the results on a benchmark demonstrate the effectiveness in comparison with other state-of-the-art algorithms.

In this study, we propose an effective and efficient algorithm for unconstrained video object segmentation, which is achieved in a Markov random field (MRF). In the MRF graph, each node is modeled as a superpixel and labeled as either foreground or background during the segmentation process. The unary potential is computed for each node by learning a transductive SVM classifier under supervision by a few labeled frames. The pairwise potential is used for the spatial-temporal smoothness. In addition, a high-order potential based on the multinomial event model is employed to enhance the appearance consistency throughout the frames. To minimize this intractable feature, we also introduce a more efficient technique that simply extends the original MRF structure. The proposed approach was evaluated in experiments with different measures and the results based on a benchmark demonstrated its effectiveness compared with other state-of-the-art algorithms.

We propose a new photo-guided drawing tool for the easy generation of artistic strokes in the style of flower painting, which is the most representative one in Chinese paintings, through stylizing rough sketch lines over a photoimage. While existing style migration methods aim at processing the entire image, we depict that a stroke-based stylization framework better performs detailed style features and catches the user’s intent. The steps in our technique are first to correct user inaccurate strokes that mark candidates of style patterns on a painting and skeletal paths to be stylized over a photograph, then to select the optimal pattern for every skeletal stroke, and finally, to map the pattern onto skeletal stroke by shape calculation and texture synthesis. This tool can tolerate errors from user inputs and work well for touch devices. Our experiments prove that the paintings our technique helped to generate show close to the specified style and are aesthetically satisfied by normal users.

Automating the production of 2D hand-drawn animations is a significant and interesting component in computer graphics and vision. However, traditional methods in animation production pipeline always use physically or geometrically based models which are consuming due to complicated and massive computations, reducing their practicability. In this paper, we propose an efficient data-driven approach to create hand-drawn animations in an automatic manner. The key idea is to employ a correspondence match-based random search process to extract the geometry samples and the global motion pattern in an input animation sequence and then to generate a new output sequence through a coarse-to-fine sample-based synthesis algorithm. Our experiments demonstrate that our method achieves good results with high quality and performance, producing a range of artistic effects that previously required disparate and professional techniques.

This paper introduces versatile filters to construct efficient convolutional neural networks that are widely used in various visual recognition tasks. Considering the demands of efficient deep learning techniques running on cost-effective hardware, a number of methods have been developed to learn compact neural networks. Most of these works aim to slim down filters in different ways, \eg,~investigating small, sparse or quantized filters. In contrast, we treat filters from an additive perspective. A series of secondary filters can be derived from a primary filter with the help of binary masks. These secondary filters all inherit in the primary filter without occupying more storage, but once been unfolded in computation they could significantly enhance the capability of the filter by integrating information extracted from different receptive fields. Besides spatial versatile filters, we additionally investigate versatile filters from the channel perspective. Binary masks can be further customized for different primary filters under orthogonal constraints. We conduct theoretical analysis on network complexity and an efficient convolution scheme is introduced. Experimental results on benchmark datasets and neural networks demonstrate that our versatile filters are able to achieve comparable accuracy as that of original filters, but require less memory and computation cost., Comment: Accepted by TPAMI. Extended version of NeurIPS 2018 paper

Internet video delivery has undergone a tremendous explosion of growth over the past few years. However, the quality of video delivery system greatly depends on the Internet bandwidth. Deep Neural Networks (DNNs) are utilized to improve the quality of video delivery recently. These methods divide a video into chunks, and stream LR video chunks and corresponding content-aware models to the client. The client runs the inference of models to super-resolve the LR chunks. Consequently, a large number of models are streamed in order to deliver a video. In this paper, we first carefully study the relation between models of different chunks, then we tactfully design a joint training framework along with the Content-aware Feature Modulation (CaFM) layer to compress these models for neural video delivery. {\bf With our method, each video chunk only requires less than $1\% $ of original parameters to be streamed, achieving even better SR performance.} We conduct extensive experiments across various SR backbones, video time length, and scaling factors to demonstrate the advantages of our method. Besides, our method can be also viewed as a new approach of video coding. Our primary experiments achieve better video quality compared with the commercial H.264 and H.265 standard under the same storage cost, showing the great potential of the proposed method. Code is available at:\url{https://github.com/Neural-video-delivery/CaFM-Pytorch-ICCV2021}, Comment: Accepted by ICCV 2021

Unified simulation of versatile elastoplastic materials and different dimensions offers many advantages in animation production, contact handling, and hardware acceleration. The unstructured particle representation is particularly suitable for this task, thanks to its simplicity. However, previous meshless techniques either need too much computational cost for addressing stability issues, or lack physical meanings and fail to generate interesting deformation behaviors, such as the Poisson effect. In this paper, we study the development of an elastoplastic model under the state-based peridynamics framework, which uses integrals rather than partial derivatives in its formulation. To model elasticity, we propose a unique constitutive model and an efficient iterative simulator solved in a projective dynamics way. To handle plastic behaviors, we incorporate our simulator with the Drucker-Prager yield criterion and a reference position update scheme, both of which are implemented under peridynamics. Finally, we show how to strengthen the simulator by position-based constraints and spatially varying stiffness models, to achieve incompressibility, particle redistribution, cohesion, and friction effects in viscoelastic and granular flows. Our experiments demonstrate that our unified, meshless simulator is flexible, efficient, robust, and friendly with parallel computing.

Constructive Solid Geometry (CSG) is one of the popular techniques that is widely applied in 3D modeling. It combines primitive solids using Boolean operations. However, the trade-off between efficiency and robustness of Boolean evaluation is difficult to balance. Previous methods sacrifice either efficiency or robustness to achieve advantages in one perspective. Recent works attempt to achieve excellent performance in both aspects through replacing the conventional vertex-based representations (V-reps) with plane-based representations (P-reps) of polyhedrons. Different from V-reps, the P-reps use plane coefficients as meta-data and can lead to benign robustness. However, methods using P-reps have disadvantages in efficiency compared to methods using V-reps. In this paper, we proposed a Boolean evaluation approach that absorbs both the efficiency of V-reps based methods and robustness of P-reps based methods. We design a Boolean evaluation method combining P-reps with V-reps. The P-reps information is utilized for exact predicate computation while information in V-reps is collected for fast topology query and coarse tests. Our proposed approach is variadic: it evaluates a Boolean expression regarding multi-input meshes as a whole rather than a tree of decomposed binary operations. We conduct massive experiments and compare our results with those generated by the state-of-the-art methods. Experimental results show that our approach is robust for solid inputs and has advantages in performance compared to some previous non-robust methods.

We propose an efficient non-incremental approach to evaluate the boundary of constructive solid geometry (CSG) in this paper. In existing CSG evaluation methods, the face membership classification is a bottleneck in executive efficiency. To increase the executive speed, we take advantages of local coherence of space labels to accelerate the classification process. We designed a two-level grouping scheme to group faces that share specific space labels to reduce redundant computation. To further enhance the performance of our approach in the non-incremental evaluation, we optimize our model generation which can produce the results in one-shot without performing a step-by-step evaluation of the Boolean operations. The robustness of our approach is strengthened by the plane-based geometry embedded in the intersection computation. Various experiments in comparison with state-of-the-art techniques have shown that our approach outperforms previous methods in boundary evaluation of both trivial and complicated CSG with massive faces while maintaining high robustness.

Recently, single-image rain removal has attracted sustained attention. In this paper, we utilize a simple shape prior, the fact that the rain streaks often appear in an elongated elliptical shape, in a particular way for single image de-raining. With this prior, the patches of the rain layer are forced to have a high correlation with our synthetic ellipse matrix. Then, the rain layer is refined during each layer decomposition process with the correlation constraints. The experiments show that the proposed method outperforms the existing single image de-raining methods on both real and synthetic data. Moreover, our approach runs much faster than the other state-of-the-art de-raining methods.

Non-uniform motion blur due to object movement or camera jitter is a common phenomenon in videos. However, the state-of-the-art video deblurring methods used to deal with this problem can introduce artifacts, and may sometimes fail to handle motion blur due to the movements of the object or the camera. In this paper, we propose a non-uniform motion model to deblur video frames. The proposed method is based on superpixel matching in the video sequence to reconstruct sharp frames from blurry ones. To identify a suitable sharp superpixel to replace a blurry one, we enrich the search space with a non-uniform motion blur kernel, and use a generalized PatchMatch algorithm to handle rotation, scale, and blur differences in the matching step. Instead of using pixel-based or regular patch-based representation, we adopt a superpixel-based representation, and use color and motion to gather similar pixels. Our non-uniform motion blur kernels are estimated from the motion field of these superpixels, and our spatially varying motion model considers spatial and temporal coherence to find sharp superpixels. Experimental results showed that the proposed method can reconstruct sharp video frames from blurred frames caused by complex object and camera movements, and performs better than the state-of-the-art methods.

The central building block of convolutional neural networks (CNNs) is the convolution operator, which enables networks to construct informative features by fusing both spatial and channel-wise information within local receptive fields at each layer. A broad range of prior research has investigated the spatial component of this relationship, seeking to strengthen the representational power of a CNN by enhancing the quality of spatial encodings throughout its feature hierarchy. In this work, we focus instead on the channel relationship and propose a novel architectural unit, which we term the "Squeeze-and-Excitation" (SE) block, that adaptively recalibrates channel-wise feature responses by explicitly modelling interdependencies between channels. We show that these blocks can be stacked together to form SENet architectures that generalise extremely effectively across different datasets. We further demonstrate that SE blocks bring significant improvements in performance for existing state-of-the-art CNNs at slight additional computational cost. Squeeze-and-Excitation Networks formed the foundation of our ILSVRC 2017 classification submission which won first place and reduced the top-5 error to 2.251%, surpassing the winning entry of 2016 by a relative improvement of ~25%. Models and code are available at https://github.com/hujie-frank/SENet., journal version of the CVPR 2018 paper, accepted by TPAMI

Interactive isosurface visualisation has been made possible by mapping algorithms to GPU architectures. However, current state-of-the-art isosurfacing algorithms usually consume large amounts of GPU memory owing to the additional acceleration structures they require. As a result, the continued limitations on available GPU memory mean that they are unable to deal with the larger datasets that are now increasingly becoming prevalent. This paper proposes a new parallel isosurface-extraction algorithm that exploits the blocked organisation of the parallel threads found in modern many-core platforms to achieve fast isosurface extraction and reduce the associated memory requirements. This is achieved by optimising thread co-operation within thread-blocks and reducing redundant computation; ultimately, an indexed triangular mesh can be produced. Experiments have shown that the proposed algorithm is much faster (up to 10X) than state-of-the-art GPU algorithms and has a much smaller memory footprint, enabling it to handle much larger datasets (up to 64X) on the same GPU.

Simulating detailed dynamic hairs in real time is a challenging problem. Existing methods either simplify the strand dynamics or reduce the degrees of freedom at the cost of rich motion details. We present a real-time simulation for animating hair with high fidelity details. Our approach efficiently captures the inextensibility, bending and torsion strand mechanics, while presenting the stiction/repulsion and detailed real-time collision effects. To efficiently capture self-interactions, we factorize the phenomenon into a coarse, globally coupled volumetric, and detailed collision view. The coarse behaviors are solved with an Eulerian method via position-based density control, while detailed collisions are efficiently handled with temporal coherent link updates. We further provide a fast implicit integration via heptadiagonal matrix decomposition, which provides two to three orders of magnitude of acceleration to traditional methods. The efficiency and effectiveness of our method is validated by simulating variant motions of hair in various styles.

W. Wu · E. Wu University of Macau, Macao, China Fig. 1 Comparison of different values of . From left to right, the first one is semi-Lagrangian advection while the others used of x , 3 x and 5 x for DBCMmethod. Smaller led to more detailed and turbulent surface. can be used for controlling the details. As level set was used for identifying the fluid domain, the underlying simulations could be slightly different with different values of . Even with high tolerance value, the result was still much better than that of semi-Lagrangian advection

This paper presents a non-photorealistic rendering technique for stylizing a photograph in the line drawing style. We first construct a smooth and direction-enhancing edge flow field from the eigenvectors of the smoothed structure tensor, and then we use the flow field to guide the line drawing process. In particular, we develop a new operator for detecting step edges, which outperforms the existing edge detectors in terms of feature preservation and edge localization. Our approach works by applying the proposed detector in the direction perpendicular to the edge flow tangent and then smoothing the intermediate results along the edge flow curve. Optionally, an anisotropic nonlinear filter with an elliptical kernel is incorporated into the algorithms to extract the line edges, which may extend our technique further for creating an image to convey a hand-painting style. The presented algorithms are all highly parallel, allowing a real-time performance with GPU implementation. Experimental results show that our approach can produce more attractive and impressive line illustrations with a variety of photographs.

This paper presents a traffic-oriented modeling and simulation framework. With this tool, users can create a traffic scene quickly and easily, and then drive the vehicle in the created scene with physics simulated. Especially, the vehicle could be driven with vision technology, which is equipped with a webcam and a paper printed with controlling markers. And the paper is used as the virtual wheel to control the motion of the vehicle. At such low cost, many people can enjoy the driving fun; it also provides an easy and interesting way for driving video games with promising business values.

In this paper, we propose an efficient method to generate optimal multiresolution models for deforming mesh sequence. By this method we define the importance degree of a triangle to be embedded into the quadric error metric (QEM) as a weight parameter, so that the new metric cannot only measure distance error but also reflect geometric variation of local areas. We append a deformation degree weight to the aggregated edge contraction cost for the whole animation. This new metric can preserve not only the model's individual geometric features, but also the features only appeared during the deforming animation. In addition, a mesh optimization method has also been proposed to allow visual distortion of the simplified models to be further alleviated. Our approach is efficient in operation, easy for implementation, and as a result a good quality of dynamic approximations with well-preserved fine details can be generated at any given frame on average.

In this paper, we propose a novel method that represents the highly complex point sets by clustering and reconstructing the points to normal-mapped B-spline surfaces (NBSs). The main idea is to construct elaborate normal maps on simple surfaces for the realistic rendering of complex point-set models. Based on this observation, we developed the coarse, NBSs to approximate the original point datasets with fine surface details. In our algorithm, a genetic clustering algorithm is proposed to automatically segment the point samples into several clusters according to their statistical properties, and a network of B-spline patches with normal maps are constructed according to the clustering results. In respect to signal processing, our algorithm decouples the original data into two collections: coarse geometry information and fine surface details. The multi-level B-spline surfaces are employed to describe the coarse geometry information, and the normal maps are constructed to capture the fine surface details. Our experimental results show that this representation facilitates the modeling and rendering of complex point sets without losing the visual qualities.

Due to high computational expense of simulating the physical model of atmospheric effect, rendering realistic natural light phenomena in real time is a challenging work. In this paper, we present a new approach for dynamic light shaft construction and rendering. Unified volumes construction is proposed for both dynamic light shaft and shadow, which is embedded into an efficient method of rendering natural light shafts with atmospheric scattering. We present the analytic formula of scattering computation without numerical integration and its implementation on current GPU. Our approach can simulate not only the lighting effect with single light source but also multiple parallel light sources according to the physical model of skylight and sunlight.

The paper presents a new algorithm for point-in-polygon queries. By decomposing a polygon into a set of convex polygons and then constructing a BSP tree, the algorithm performs an inclusion query against the polygon in two steps. In the first step, it finds the convex polygon that is most likely to contain the query point through the BSP tree, and then in the second step, it tests whether the query point is in the convex polygon by employing advanced techniques for point-in-convex-polygon queries. The new algorithm is immune of singularity and has a performance of conducting a query by time demand in O(logn), and storage requirement in O(n). The time complexity of the preprocessing is O(nlogn) for convex decomposition and BSP tree construction. The new algorithm has its performance in various aspects comparable to the state-of-art algorithms based on trapezoidation. Furthermore, as the number of convex polygons is always much fewer than the number of trapezoids in partitioning a polygon, and a high-quality search tree can always be constructed to manage the convex polygons, the new algorithm can run much faster than the trapezoidation-based algorithms, up to over double folds in maximum.

The progressive mesh representation is in general introduced for a continuous level of detail, and with it smooth transition between meshes can be achieved. However, the method is impractical for realtime simplification. Despite that, the realtime generation of progressive meshes is often required for applications such as in a robot simulation system or a flight simulation system where models are constantly modified by the actions of object collision, object destruction and reconstruction, etc. A new method is presented in this paper to tackle realtime simplification through progressive meshes. By the algorithm only the current mesh and readily accessible local information of each vertex, such as vertex position, edge length, neighboring vertices, adjacent faces, and face normals, are considered at each simplification step. All edge-collapse costs are calculated and sorted into a binary tree using the heap sort algorithm at the initial stage. At the iterative stages only the neighboring affected costs need be recalculated and rearranged in the binary tree. Other properties, such as vertex color and face texture, are processed in the same way as the geometry. The algorithm greatly improves the time performance under the constraint that the quality of the generated meshes be acceptable. Tests show that the algorithm is viable in a realtime simplification for medium-scale virtual models on PC platforms.

International audience; In fast figure animation, motion blur is of crucial importance, and this is especially true when an artist wants to generate exaggerating effect through figure motion. For a quite long period of time, animators seek the answer by using certain kind of image blending, no matter by the means of hardware or software. In recent years, methods based on 3D geometry of the motion figure with global illumination become gradually in demand, as they could deliver relatively high quality of motion blur effect. However, the computation cost in those methods is always very high, thus real time rendering become quite difficult to achieve. In this paper, a real-time motion effect based on 3D geometric approach is proposed, in which a special effect along the motion trajectory based on fluid simulation is combined with the volumetric motion blur. Furthermore, the motion trajectory would be decomposed and multi-pass geometry rendering would be employed to achieve geometry instancing for reuse. In this manner, the redundant calculation of each frame could be avoided, and the limitation of trajectory generation would be broken. In the pipeline, we separate motion tracking and fluid solution, to support various fluid effects flexibly. The scheme we present makes use of GPU geometry shading in parallel, aiming at guaranteeing high efficiency of computation while delivering splendid rendering. As a result, real time rendering including the motion blur effect is achieved.

International audience; With regard to skeletal character animation, the question on how to ensure the quality of transition between key frames is of crucial importance. The lack of properly defined motion ranges based on movements would leave the animator with no choice but intervene the result based on camera perspective afterwards, thus creating a lot more work for the animator to modify or clean up the animation curves. Although a number of methods have been raised in these years, such as the Linear Blending Skinning (LBS), they may still have shortcomings in some specific cases, one of which is the obvious unnatural deformation around the joint areas. The primary investigation in this paper is directed to address the problem and improve the framework of rendering in real-time environment with satisfactory skinning effect to the aforementioned scenario, with assistance of GPU computation

International audience; The sketch-based shape modeling is one of the most challenging and active problems in computer graphics. In this paper, we present an interactive modeling system for generating free-form surfaces using a 2D sketch interface. Since inferring 3D shape from 2D sketches is an one to many function with no unique solution, we propose to interpret the given 2D curve to be the projection of the 3D curve that has minimum curvature among all the candidates in 3D. In this way, firstly, we present an algorithm to efficiently find a close approximation of this minimum curvature 3D space curve. In the second step, our system could identify the 3D surfaces automatically, and then we apply Delaunay triangulation on these surfaces. Finally, the shape of the triangular surface mesh that follows the 3D profile curves is computed using harmonic interpolation by solving Laplacian equations. We present experimental results on various kinds of drawings by the interactive modeler

Background prior has been widely used in many salient object detection models with promising results. These methods assume that the image boundary is all background. Then, color feature based methods are used to extract the salient object. However, such assumption may be inaccurate when the salient object is partially cropped by the image boundary. Besides, using only color feature is also insufficient. We present a novel salient object detection model based on background selection and multi-features. Firstly, we present a simple but effective method to pick out more reliable background seeds. Secondly, we utilize multi-features enhanced graph-based manifold ranking to get the saliency maps. Finally, we also present the salient object segmentation via computed saliency map. Qualitative and quantitative evaluation results on three widely used data sets demonstrate significant appeal and advantages of our technique compared with many state-of-the art models.

Relief texture mapping is an image-based rendering technique which can successfully support the representation of 3D surface details and view motion parallax. It has the potential to significantly increase visual realism of rendered geometry while keeping system load constant. In this paper, FPGA (Field Programmable Gate Array) chip technology is applied to this three-dimensional image warping method. A relief texture mapping system has been implemented on a reprogrammable and reconfigurable FPGA board. The algorithm is optimized for the specific architecture and the framework is customized for circuit resources, which can be flexibly changed for other structures. In our design, we take advantage of inherent parallelism of the algorithm by concatenating multiple warping engines and well organizing data in memory space. Experimental results show high image quality with improved rendering speed.

This paper presents a new data structure, called stick textures, to give a continuous and compact 3D representation of the scene displayed in depth images. A stick texture is constructed for each pixel of a depth image, composed of the interpolated pixels to fill the gaps between the pixel and its neighbour pixels covered by a same continuous surface. Thus, rendering sticks instead of pixels can save operations on hole-filling for acceleration, and produce high quality novel images with 3D details and view motion parallax displayed very well. In addition, stick textures are always short, and every stick texture can be rendered as a pixel. Therefore, stick textures have not a high storage requirement, and they can be easily integrated with existing image-based rendering methods to achieve high rendering efficiency, e.g., using GPUs for acceleration and conveniently handling large-scale scenes, even the scenes with multi-layered objects. Experimental results demonstrate the advantages of stick textures in image quality, rendering speed and storage requirement.

The 2D point-in-polygon test is a fundamental problem in geometry, and of importance in various applications in computer graphics and other areas. In taking advantage of the basic idea of the polygon scan conversion algorithm, a novel method for the point-in-polygon test is proposed in this paper, capable of handling simple polygons in arbitrary shapes, possibly with holes. In the preprocess of the method, the edges of the polygon are classified into layers according to the occlusion relations between the edges viewed orthogonally in a direction, called the test direction, which guarantees that the edges of a layer can be occluded by only the edges of its preceding layers. At the same time, the edges at each layer are queued up respectively along the direction vertical to the test direction because there is no occlusion relation between the edges of the same layer. As a result, based on the layers, the calculation of the segments by the line through the tested point to intersect the polygon along the test direction, and then the inclusion test of the point against the segments could be feasibly made. The method has a low storage requirement in O(n), here, n is the number of the edges of the polygon. The time complexity of its preprocess ranges from O(n) to O(n^2), depending on the polygon shape and the test direction. And its inclusion test has a time complexity between O(log(n)) and O(n), but less than O((log(n))^2) in most cases, depending on the construction of the layers. In the case of convex polygons and monotone polygons, the time complexity for the preprocess and the inclusion test could be O(n) and O(log(n)), respectively. On the other hand, the method is also easy to integrate with a variety of existing methods such as ray crossing methods and grid-based methods for improving the inclusion test further. Experimental results show that the method is robust and efficient in computation.

In volume rendering, an important issue in acceleration is to reduce the calculations for occluded voxels. Although this issue has been addressed in the ray casting approach, it is difficult to apply the idea to the projection approach due to uncertain termination conditions. In this paper, we propose a new method to effectively address the exclusion problem in the projection approach, so the rendering process can be accelerated without impairing the rendered image quality. In the rendering process, this new method employs the dynamic screen technique to manage the pixels whose accumulated opacity has not reached 1.0. A ray-cast link at each pixel is set up to record the rendered voxels for the corresponding ray cast from the pixel to intersect. According to the rendered voxels covering the pixels whose accumulated opacity value is below 1.0, visible voxels are selected to render from front to back by the neighboring relationship between the rendered voxels and the voxels to be rendered. Thus, the occluded voxels are dynamically excluded from the loading and rendering processes accurately. Our proposed method can be in general applied to both parallel and perspective projections, using regular and irregular volume datasets. Our experimental results showed that the proposed method can significantly accelerate volume rendering if the data volume has a high percentage of occluded voxels. This method can also perform fairly efficiently for the expensive shading calculations if requested in volume rendering.

Taking advantage of the parallelism and programmability of GPU, we solve the fluid dynamics problem completely on GPU. Different from previous methods, the whole computation is accelerated in our method by packing the scalar and vector variables into four channels of texels. In order to be adaptive to the arbitrary boundary conditions, we group the grid nodes into different types according to their positions relative to obstacles and search the node that determines the value of the current node. Then we compute the texture coordinates offsets according to the type of the boundary condition of each node to determine the corresponding variables and achieve the interaction of flows with obstacles set freely by users. The test results prove the efficiency of our method and exhibit the potential of GPU for general-purpose computations. Copyright © 2004 John Wiley & Sons, Ltd.

The algorithm efficiency of existing adaptable out-of-core simplification algorithms is relatively low because these methods often require scanning the original model over one time. In this paper, we present an adaptive clustering method, known as the Balanced Tiling (BT) method, for out-of-core simplification which only needs one pass over the input model. The key idea behind BT is that the model surface can be recorded in a memory efficient manner using surface coding and the global distribution of surface details can be obtained through quadric quantizing of the original model. Besides the feature edge areas, our algorithm can position the corner areas, which are not noticed by some other out-of-core simplification approaches such as the Fei's BR algorithm, by analyzing the Garland's quadric error matrix. Based on this analysis, adaptive clustering is achieved by restoring detail areas with cluster split operations while further simplifying smooth areas with edge collapse operations. BT is especially suitable for handling super huge meshes because the I/O time is greatly reduced. BT produces high quality approximations and the memory requirement is small, being only related with the size of the output model.

Author(s): Wang, Wencheng; Wu, En-Hua; Max, Nelson | Abstract: Selective visualization is a solution for visualizing data of large size and dimensionality. In this paper a new method is proposed for effectively rendering certain chosen parts among the full set of data in terms of a colour buffer, referred to as the virtual plane, for storing intermediate results. By this method, scientists may concentrate their attention on the contents of data in which they are interested. Besides, the method could be easily integrated with all the current direct volume rendering techniques, especially progressive refinement methods and selective methods.