• 한국통신학회논문지
• /
• 제25권9B호
• /
• pp.1604-1608
• /
• 2000

The three-step search (TSS) algorithm for block motion estimation has been widely used in real-time video coding due to the simplicity of the algorithm significant reduction of computationl cost and good performance. In this paper an improved three-step search (ITS) algorithm is proposed to improve the performance of the TSS algorithm. Simulation results show that in terms of motion compensation errors the proposed ITSS outperforms some popular fast search algorithms while it has the lower computational complexity.

• Communications for Statistical Applications and Methods
• /
• 제15권1호
• /
• pp.1-11
• /
• 2008

혼합정규분포로부터 얻은 자료의 크기가 크면 EM 알고리즘으로 모수를 추정하는 경우 추정에 많은 시간이 걸리며 이는 실시간 음성인식 분야등에서는 적용이 어렵게 되는 문제가 발생한다. 대용량 자료를 구간도수로 요약하여 구간도수 EM 알고리즘을 적용하면 표준 EM 알고리즘에 비해 실행속도가 획기적으로 개선되며 더욱이 구간도수 EM 알고리즘에서의 추정치의 효율성이 표준 EM 알고리즘에 근접함을 시뮬레이션 실험을 통하여 보였다.

• 방송공학회논문지
• /
• 제16권6호
• /
• pp.902-915
• /
• 2011

영역기반 스테레오 매칭 분야에서 최근 인간의 시각체계(Human Visual System)에 기반하여 영역내의 밝기값과 거리값에 따라 적응적으로 가중치를 부여하는 적응적 영역 가중치(Adaptive Support-Weight, ASW) 방법이 좋은 매칭 결과를 보이고 있다. 그러나 이 방법은 좋은 매칭 결과에 비해서 많은 연산비용을 필요로 하게 되고, 매칭의 실시간 시스템화에 큰 장애요소로 작용한다. 이에 Bilateral filter 수식으로 근사화 후 Integral Histogram 기법을 적용하여 영역 윈도우의 크기에 상관없이 상수 시간 O(1) 내에 매칭을 수행하는 연구가 진행되었다. 하지만 이 방법은 근사화 과정에서의 원 ASW 수식을 왜곡하기 때문에 매칭 정확도의 손실을 가져오게 된다. 본 논문에서는 적응적 영역 가중치 알고리즘의 매칭 정확도를 유지하면서 적응적 영역 가중치 알고리즘의 계산 비용을 줄이는 알고리즘을 제안한다. 이를 위해 영역내의 픽셀을 그룹화하여 근사화된 매칭을 수행하는 Sub-Block 방법과 영상의 에지 정보에 따라 적응적으로 시차 탐색 범위를 조정하는 방법을 제안한다. 결과적으로 제안된 기법은 기존 방식보다 좋은 매칭 정확도를 유지하면서도 효율적으로 계산 수행 시간을 줄이게 된다.

• Journal of the Korean Society for Industrial and Applied Mathematics
• /
• 제20권3호
• /
• pp.175-202
• /
• 2016

Ultrasound imaging is a widely used tool for visualizing human body's internal organs and quantifying clinical parameters. Due to its advantages such as safety, non-invasiveness, portability, low cost and real-time 2D/3D imaging, diagnostic ultrasound industry has steadily grown. Since the technology advancements such as digital beam-forming, Doppler ultrasound, real-time 3D imaging and automated diagnosis techniques, there are still a lot of demands for image quality improvement, faster and accurate imaging, 3D color Doppler imaging and advanced functional imaging modes. In order to satisfy those demands, mathematics should be used properly and effectively in ultrasound imaging. Mathematics has been used commonly as mathematical modelling, numerical solutions and visualization, combined with science and engineering. In this article, we describe a brief history of ultrasound imaging, its basic principle, its applications in obstetrics/gynecology, cardiology and radiology, domestic-industrial products, contributions of mathematics and challenging issues in ultrasound imaging.

• 대한전자공학회:학술대회논문집
• /
• 대한전자공학회 2006년도 하계종합학술대회
• /
• pp.873-874
• /
• 2006

In this paper, we have developed a ventricular fibrillation detection system based on DSP processor. The developed system was able to detect VF in real time correctly and quickly. We compared the performance of the floating point simulation with that of fixed point simulation. The computational cost of fixed point simulation was remarkably reduced than that of floating point simulation.

• Journal of Communications and Networks
• /
• 제12권3호
• /
• pp.253-265
• /
• 2010

The problem of pricing for a telecommunication network is investigated with respect to the users' sensitivity to the pricing structure. A functional optimization problem is formulated, in order to compute price reallocations as functions of data collected in real time during the network evolution. No a-priori knowledge about the users' utility functions and the traffic demands is required, since adaptive reactions to the network conditions are sought in real time. To this aim, a neural approximation technique is studied to exploit an optimal pricing control law, able to counteract traffic changes with a small on-line computational effort. Owing to the generality of the mathematical framework under investigation, our control methodology can be generalized for other decision variables and cost functionals.

• 한국멀티미디어학회논문지
• /
• 제22권2호
• /
• pp.311-321
• /
• 2019

Virtual reality applications in Head-Mounted Displays require high frame rates and low latency rendering techniques. Ray tracing offers many benefits, such as high-quality image generation, but has not been utilized due to lower performance than rasterization. But that can obtain good result combined with gaze-tracking technology and human visual system's operation principle. In this paper, we propose a method to optimize the foveated sampling map and to maintain the visual quality through the fast voronoi nearest interpolation. The proposed method further reduces the computational cost that has been improved by the previous foveated sampling. It also smoothes the voronoi boundary using adaptive sibson interpolation, which was not possible in real-time. As a result, the proposed method can render real-time high-quality images with low visual difference.

• 한국멀티미디어학회논문지
• /
• 제22권2호
• /
• pp.167-177
• /
• 2019

Even though so much progresses have been achieved in Multiple Object Tracking (MOT), most of reported MOT methods are not still satisfactory for commercial embedded products like Pan-Tilt-Zoom (PTZ) camera. In this paper, we propose a real-time multiple pedestrians tracking method for embedded environments. First, we design a new light weight convolutional neural network(CNN)-based pedestrian detector, which is constructed to detect even small size pedestrians, as well. For further saving of processing time, the designed detector is applied for every other frame, and Kalman filter is employed to predict pedestrians' positions in frames where the designed CNN-based detector is not applied. The pose orientation information is incorporated to enhance object association for tracking pedestrians without further computational cost. Through experiments on Nvidia's embedded computing board, Jetson TX2, it is verified that the designed pedestrian detector detects even small size pedestrians fast and well, compared to many state-of-the-art detectors, and that the proposed tracking method can track pedestrians in real-time and show accuracy performance comparably to performances of many state-of-the-art tracking methods, which do not target for operation in embedded systems.

• Structural Engineering and Mechanics
• /
• 제67권6호
• /
• pp.643-658
• /
• 2018

In this paper, we introduce a new framework for running the finite element (FE) packages inside an online Loop together with MATLAB. Contrary to the Hardware-in-the-Loop techniques (HiL), in the proposed Software-in-the-Loop framework (SiL), the FE package represents a simulation platform replicating the real system which can be out of access due to several strategic reasons, e.g., costs and accessibility. Practically, SiL for sophisticated structural design and multi-physical simulations provides a platform for preliminary tests before prototyping and mass production. This feature may reduce the new product's costs significantly and may add several flexibilities in implementing different instruments with the goal of shortlisting the most cost-effective ones before moving to real-time experiments for the civil and mechanical systems. The proposed SiL interconnection is not limited to ABAQUS as long as the host FE package is capable of executing user-defined commands in FORTRAN language. The focal point of this research is on using the compiled FORTRAN subroutine as a messenger between ABAQUS/CAE kernel and MATLAB Engine. In order to show the generality of the proposed scheme, the limitations of the available SiL schemes in the literature are addressed in this paper. Additionally, all technical details for establishing the connection between FEM and MATLAB are provided for the interested reader. Finally, two numerical sub-problems are defined for offline and online post-processing, i.e., offline optimization and closed-loop system performance analysis in control theory.

• 방송공학회논문지
• /
• 제27권5호
• /
• pp.812-815
• /
• 2022

햇빛 아래에서 디지털 기기에 표시되는 영상은 일반적으로 원본 영상보다 어둡게 인식되어 시인성이 저하된다. 더 나은 시인성을 위해서, 주변광에 적응적인 전역 휘도 보상 혹은 톤 매핑이 필요하다. 하지만 기존의 기법들은 색차 보상에 한계가 존재하고, 무거운 계산 비용 때문에 실제 환경에서 활용하는데 어려움이 존재한다. 이를 해결하기 위해 본 논문에서는 휘도와 색차를 모두 보상하는 Piece-wise Linear Enhancement Curves (PLECs) 기반 영상 개선 기법을 제안한다. 이때, PLECs은 딥러닝을 통해 회귀 되며, lookup table 형식으로 구현되어 실시간 동작이 가능하다. 실험 결과 제안 방법이 낮은 계산 비용으로 원본 영상 대비 더 나은 시인성을 가짐을 보인다