Earticle

효과적인 영상 복원을 달성하려면 모델이 맥락정보(contextual information)를 파악하는 것이 중요하다. 합성곱 신경망(CNN) 기반 알고리즘을 사용한 이전 연구는 장거리 종속성이 부족하여 모델이 상황별 정보를 캡처 할 수 없다는 한계에 직면했다. 이를 해결하기 위해 본 논문에서는 고품질 영상 복원을 위한 다중 차원의 윈도우 (Multi-Scale Window) 기반 비전 트랜스포머 모델을 제안한다. 다중 차원 기반 비전 트랜스포머를 도입함으로써 다양한 창 크기의 영향을 반영하고 이에 따른 맥락정보를 얻을 수 있다. 또한 손실 마스크 업데이트 모듈을 적용 하여 효율적인 연산방식을 도입하였다. 실험 결과에 기반하면 본 논문의 제안 모델은 누락된 영역을 효과적으로 복원하고 벤치마크 데이터셋에서 FID 수치가 다른 최신 복원 모델들의 결과에 비해 약 50%이상 낮은 것으로 확 인 되었다. 이는 제안 복원 모델의 우수성을 검증한다고 할 수 있다.

To achieve successful image inpainting, it is crucial for a model to understand contextual information. Previous research using Convolutional Neural Network (CNN)-based algorithms have limitations in capturing context-specific information due to a lack of long-range dependencies. To address this issue, we proposes a Multiscale Window based Vision Transformer (MWVT) model for high-quality image completion. By introducing a multi-dimensional vision transformer, the proposed MWVT can capture context information by using various window sizes, resulting in the reflectance of their respective influences. Furthermore, our MWVT incorporates a mask update module to reduce the computational cost. According to our experiments, the proposed model in this paper effectively restores the missing area and the FID value in the benchmark dataset is about 50% lower than the results of other state-of-the-art restoration models.

Low-resolution voxel models are widely used in many computer games such as CrossyRoadTM and MinecraftTM and in many block games such as LegoTM. We present a multi- scale voxelization algorithm that abstracts a complex polygonal model to a low-resolution voxel model. Our scheme is distinguished from other voxelization schemes in the point that our scheme controls the level of abstraction according to the complexity of the model. We present a two-stage algorithm: in the first stage, we build a rough voxel model, which is carved to fit the target model in the second stage. We empoly an OpenGL-based slicing algorithm to build a rough model, which is constructed in twofolds. The exterior and interior of the model is constructed separately and merged to complete the rough model. In the second stage, we compute the silhouette of the input model and carve the rough model to improve the completeness of the final result. We test our algorithm for various polygonal models including famous animation character models to prove the excellency of our scheme.

본 논문에서는 사진이나 동영상으로부터 게임 배경 영상을 생성하기 위한 영상 추상화 기법을 제시한다. DeCarlo 등이 제안한 방법 [9]에 기반하여 개발된 이 방법은 에이다부스트 알고리즘을 이용해서 인식된 여러 개의 부분과 나머지 부분들에 대해서 서로 다른 스케일로 추상화를 수행함으로써 선택적으로 다중 스케일 구조의 추상화된 영상을 생성할 수 있다. 이 방법은 여러 정지 영상 및 동영상에 적용되어서 다양한 결과를 보여준다. 이 방법은 제한된 자원을 가진 게임 개발 환경에서 게임 배경에 많은 자원을 투입하기 힘든 상황에서 유용하게 사용될 수 있다.

We present an image abstraction method for generating game background images from photographs or videos. The proposed method, which is based on [9], abstracts images into a game background image of a multi-scaled structure after selecting important parts of the image using Adaboost algorithm. We tested this method for various photographs and movie clips. This scheme can be a useful tool for the game development companies that have limited resources for building game backgrounds.

There are many mathematical models or chemical models available to describe the avascular tumor growth. However, few of them take into account the necessary physical growth for the avascular tumor. In this paper, we proposed a multiscale finite element model that spans three distinct scales for the avascular tumor growth. At the extracellular microenvironment level, we proposed a new set of piecewise reaction diffusion equations to describe the extracellular microenvironment of the tumor growth. At the intracellular cycle level, we proposed a cell cycle control factor to model the key role that played by the G1-S phase transition. At the cellular level, we introduced the proposed cell cycle control factor to combine the extracellular microenvironment, the intracellular cycle and the cellular cycle into the total energy equations to form our finite element cellular model. Experiments were performed to simulate the growth of EMT6/Ro mouse mammary cell under three different kinds of initial chemical concentrations. Results showed that the proposed approach outperformed the baseline approach in terms of tumor cell number and tumor cell volume.

An image segmentation fast method based on multi-scale belief propagation is proposed to solve the concrete CT image segmentations problem. Firstly, according to the feature of belief propagation algorithm, a self-characteristic multiscale belief propagation(MBP) is proposed; Then, according to compute complexity problem in process of belief messages propagation, a method to reduce quantity of algorithm compute is proposed; Finally, using standard images to validate nicety and speediness on our method ,and applying on concrete CT image segmentation. The experiment results show that the proposed method can improve efficiency and precision of the image segmentation, and afford an important assisting method on concrete meso-structure CT image study of architecture projection. The method has important projection applying meaning.

A new switching median filter is proposed for denoising of gray-scale images, extremely corrupted by salt-and-pepper noise. The proposed model for noise removal is a multiscale detection based adaptive median filter. This method consists of mainly two parts, namely, the thresholding based multiscale noise detection and the filtering. The detection of impulse noise is carried out in two stages. First, multiscale filtering of the corrupted image is carried out using Gaussian kernels at different scales and errors between the original and the filtered images at different scales are obtained. In the next stage, the errors at different scales are added and then thresholded to detect the impulse noise. The filtering of impulses, detected in the first stage of the proposed filter, is finally carried out using an adaptive median filter. Incorporation of a multiscale method into the noise detection stage followed by thresholding has led to more reliable and efficient impulse noise detection, especially, at high noise ratios. To validate the efficacy of proposed scheme, extensive simulations and comparisons are done with the competent schemes under a wide range (10% to 90%) of noise densities. The results show that the proposed scheme works much better in suppressing high level noise than other schemes, keeping the edges and fine details of the original image almost intact.

Carbon nanotubes (CNTs) are being used extensively as reinforcing materials in polymer matrix composites because of their high strength, stiffness and resilience, as well as superior mechanical, electrical and thermal properties. Incorporating CNTs in polymer matrix composites can potentially enhance the strength and stiffness of composites significantly. In this paper, the effective elastic properties of nanocomposites (CNTs/Epoxy) at different volume fractions of CNTs and multiscale composites (Glass/CNTs/Epoxy) at 5% volume fraction of CNTs are evaluated using finite element method (FEM). 3-D finite element models using square representative volume element (RVE) incorporating necessary boundary conditions are developed. For validity the obtained results are compared with that of classical theories of equivalent material properties. Good agreement between them has been observed. Further the effect of CNT-integration in fiber-reinforced composites (three-phase) is also studied.

Fast multiscale directional filter bank (FMDFB) is a multiscale image representation scheme that is suitable for several image processing applications such as image denoising, texture retrieval, and pattern recognition. It is constructed using the iterative filter structures namely Laplacian pyramid and directional filter bank. In this paper, we present the method for designing the FIR filters required to construct FMDFB. Also, we demonstrate the perfect reconstruction and critically sampled property of FMDFB.

In this paper, we propose an analysis method of spatial and temporal characteristics of education network traffic based on flexible multiscale entropy (FMSE). As an improved method, flexible multiscale entropy has a significant improvement in stability and accuracy over multiscale entropy (MSE). We analyze network traffic in different time scales, space scales and traffic levels by building network traffic time-space analysis model and using flexible multiscale entropy as a method to quantify complexity of different network traffic subsequences. The results show that there are distinct characteristics in the complexity of different levels of network traffic. We also find that the existence of a large number of small network traffic flows has a significant influence on the complexity of network traffic.

이 논문은 충칭모델 형성에 대한 기존 연구가 중앙정부와 공산당의 역할을 과도하게 강조하고 다른 스케일 행위자들의 역할을 소홀히 한 것에 대한 비판적 관점에서 시작되었다. 우선 이 논문은 개혁기 중국에서 발전전략으로서의 중국모델이 문제점을 드러내자 이를 대체하기 위한 모델의 실험이 충칭 량장신구에서 수행되었다고 전제한다. 이 논문의 목적은 다중스케일 분석의 관점에서 량장신구형성에 개입한 다양한 공간스케일의 행위자와 조건들을 규명하고, 량장신구 형성과정에서 이들 사이의 역동적 상호작용을 분석하는 것이다. 이러한 목적을 달성하기 위해서 이 논문은 글로벌, 국가, 지역, 도시로 스케일을 분류하고, 각 스케일에 연결되는 다중스케일적 행위자 황치판에 주목했다. 연구 결과 황치판은 다중스케일적 행위자로서 자신의 푸동신구 경험, 중공중앙 정치국 상무위원들과의 인적 네트워크, 정책생산 능력에 기초해서, 국가스케일을 매개로 두 도시스케일을연결해 량장신구 형성에 기여했다는 결론에 도달했다.

This paper begins with criticism on preceding research of the making ofChongqing model, which excessively emphasizes the role of central government andcommunist party and neglects the role of other scale agents in the making. Thispapers assumes that when China model which had been a core development modelstarted to expose problems, the experimentation to find a new model to replacethe old model has been conducted in Liangjiang New Area(L.N.A.), Chongqing. And this paper aims to clarify multiscale agents and conditions that played an importantrole in the making of L.N.A. and to analyze dynamic interaction amongthem. To accomplish this aim, this paper classifies global, national, regional, urbanscales and pays attention to a multiscalar agent, Huang Qifan. Lastly, this paperconcludes that based on his experience in Pudong New Area, personal network withpolitburo standing committee members, and ability to produce policies, Huang Qifancontributed to the making of Liangjiang New Area connecting two urban scalesthrough national scale.

The study of applying wavelet transform in power cable system fault location has been recognized by many researchers and investigated. For performance of fault location, the fault generated transients can be captured at one end of the cable or both ends. Between two approaches, single-ended approach is less expensive and more reliable as it doesn't need communication link between the ends of the cable. So, we performs the approach based on the one. In this paper, we are going to introduce a new algorithm to discriminate the transient and the reflected signal using wavelet coefficient. For wavelet transform, the stationary wavelet transform(SWT) is applied instead of conventional DWT because SWT has redundancy properties which is more useful in noisy signal processing.

Understanding the classification of malocclusion is a crucial issue in Orthodontics. It can also help us to diagnose, treat, and understand malocclusion to establish a standard for definite class of patients. Principal component analysis (PCA) and k-means algorithms have been emerging as data analytic methods for cephalometric measurements, due to their intuitive concepts and application potentials. This study analyzed the macro- and meso-scale classification structure and feature basis vectors of 1020 (415 male, 605 female; mean age, 25 years) orthodontic patients using statistical preprocessing, PCA, random matrix theory (RMT) and k-means algorithms. RMT results show that 7 principal components (PCs) are significant standard in the extraction of features. Using k-means algorithms, 3 and 6 clusters were identified and the axes of PC1~3 were determined to be significant for patient classification. Macro-scale classification denotes skeletal Class I, II, III and PC1 means anteroposterior discrepancy of the maxilla and mandible and mandibular position. PC2 and PC3 means vertical pattern and maxillary position respectively; they played significant roles in the meso-scale classification. In conclusion, the typical patient profile (TPP) of each class showed that the data-based classification corresponds with the clinical classification of orthodontic patients. This data-based study can provide insight into the development of new diagnostic classifications.

One of the important requirements for the application of reduced-activation ferritic/martensitic (RAFM) steel is to retain proper mechanical properties under irradiation and high-temperature conditions. To simulate the yield strength and stress-strain curve of steels during high-temperature and irradiation conditions, a multiscale simulation method consisting of both microstructure and strengthening simulations was established. The simulation results of microstructure parameters were added to a superposition strengthening model, which consisted of constitutive models of different strengthening methods. Based on the simulation results, the strength contribution for different strengthening methods at both room temperature and high-temperature conditions was analyzed. The simulation results of the yield strength in irradiation and high-temperature conditions were mainly consistent with the experimental results. The optimal application field of this multiscale model was 9Cr series (7-9 wt.%Cr) RAFM steels in a condition characterized by 0.1-5 dpa (or 0 dpa) and a temperature range of $25-500^{\circ}C$.

– In the paper, to handle the problem of traditional bag-of-features model ignoring the spatial relationship of local features in human action recognition, we proposed a Multiscale Spatial Position Coding under Locality Constraint method. Specifically, to describe this spatial relationship, we proposed a mixed feature combining motion feature and multi-spatial-scale configuration. To utilize temporal information between features, sub spatial-temporal-volumes are built. Next, the pooled features of sub-STVs are obtained via max-pooling method. In classification stage, the Locality-Constrained Group Sparse Representation is adopted to utilize the intrinsic group information of the sub-STV features. The experimental results on the KTH, Weizmann, and UCF sports datasets show that our action recognition system outperforms the classical local ST feature-based recognition systems published recently.

A formulation is proposed for modelling the process of intra-granular diffusion of fission gas during irradiation of $UO_2$ under both normal operating conditions and power transients. The concept represents a simple extension of the formulation of Speight, including an estimation of the contribution of bubble motion to fission gas diffusion. The resulting equation is formally identical to the diffusion equation adopted in most models that are based on the formulation of Speight, therefore retaining the advantages in terms of simplicity of the mathematical-numerical treatment and allowing application in integral fuel performance codes. The development of the new model proposed here relies on results obtained by means of molecular dynamics simulations as well as finite element computations. The formulation is proposed for incorporation in the TRANSURANUS fuel performance code.

Radiation effects on materials are inherently multiscale phenomena in view of the fact that various processes spanning a broad range of time and length scales are involved. A multiscale modeling approach to embrittlement of pressure vessel steels is presented here. The approach includes an investigation of the mechanisms of defect accumulation, microstructure evolution and the corresponding effects on mechanical properties. An understanding of these phenomena is required to predict the behavior of structural materials under irradiation. We used molecular dynamics (MD) simulations at an atomic scale to study the evolution of high-energy displacement cascade reactions. The MD simulations yield quantitative information on primary damage. Using a database of displacement cascades generated by the MD simulations, we can estimate the accumulation of defects over diffusional length and time scales by applying kinetic Monte Carlo simulations. The evolution of the local microstructure under irradiation is responsible for changes in the physical and mechanical properties of materials. Mechanical property changes in irradiated materials are modeled by dislocation dynamics simulations, which simulate a collective motion of dislocations that interact with the defects. In this paper, we present a multi scale modeling methodology that describes reactor pressure vessel embrittlement in a light water reactor environment.