http://chineseinput.net/에서 pinyin(병음)방식으로 중국어를 변환할 수 있습니다.
변환된 중국어를 복사하여 사용하시면 됩니다.
김우석(Woosuk Kim),박병서(Byung-Seo Park),김진겸(Jin-Kyum Kim),오관정(Kwan-Jung Oh),김진웅(Jin-Woong Kim),김동욱(Dong-Wook Kim),서영호(Young-Ho Seo) 한국방송·미디어공학회 2020 방송공학회논문지 Vol.25 No.6
In this paper, we propose a method using deep learning for high-resolution display of phase holograms. If a general interpolation method is used, the brightness of the reconstruction result is lowered, and noise and afterimages occur. To solve this problem, a hologram was trained with a neural network structure that showed good performance in the single-image super resolution (SISR). As a result, it was possible to improve the problem that occurred in the reconstruction result and increase the resolution. In addition, by adjusting the number of channels to increase performance, the result increased by more than 0.3dB in same training.
정사영 벡터의 특징 분석 및 하드웨어 자원 공유기법을 이용한 저면적 Gradient Magnitude 연산 하드웨어 구현
김우석(WooSuk Kim),이주성(Juseong Lee),안호명(Ho-Myoung An) 한국정보전자통신기술학회 2016 한국정보전자통신기술학회논문지 Vol.9 No.4
본 논문은 저면적 gradient magnitude 연산을 위한 하드웨어 구조를 제안한다. 하드웨어 복잡도를 줄이기 위해 정사영 벡터의 특징 및 하드웨어 자원 공유기법을 이용했다. 제안된 하드웨어 구조는 gradient magnitude 연산 알고리즘의 변형 없이 구현되었기 때문에 gradient magnitude 데이터 품질의 열화 없이 구현될 수 있다. 제안된 저면적 gradient magnitude 연산 하드웨어는 Altera Quartus II v15.0 환경에서 Altera Cyclone VI (EP4CE115F29C7N) FPGA를 이용하여 구현되었다. 구현 결과, 기존 하드웨어 구조를 이용하여 구현한 연산기와의 비교에서 15%의 logic elements 및 38%의 embedded multiplier 절감 효과가 있음을 확인했다. In this paper, a hardware architecture of low area gradient magnitude calculator is proposed. For the hardware complexity reduction, the characteristic of orthogonal projection vector and hardware resource sharing technique are applied. The proposed hardware architecture can be implemented without degradation of the gradient magnitude data quality since the proposed hardware is implemented with original algorithm. The FPGA implementation result shows the 15% of logic elements and 38% embedded multiplier savings compared with previous work using Altera Cyclone VI (EP4CE115F29C7N) FPGA and Quartus II v15.0 environment.
딥러닝 기반의 복원 네트워크을 사용한 위상 홀로그램 비디오 압축 방법
김우석(Woosuk Kim),강지원(Ji-Won Kang),오관정(Kwan-Jung Oh),김진웅(Jin-Woong Kim),김동욱(Dong-Wook Kim),서영호(Young-Ho Seo) 한국방송·미디어공학회 2021 한국방송공학회 학술발표대회 논문집 Vol.2021 No.6
본 연구는 딥러닝 기반의 복원 모델을 사용하여, 비디오 압축을 통해 변질된 위상 홀로그램의 화질을 복원하는 방법을 제안한다. 압축 효율을 위해 위상 홀로그램의 해상도를 감소시킨 후 압축한다. 원래의 해상도로 되돌린 홀로그램을 딥러닝 모델을 사용하여 복원한다. 복원된 위상 홀로그램은 원본 홀로그램을 압축한 것보다 동일한 BPP에서 더 높은 PSNR을 보인다.
공간적 자기상관성을 고려한 자동차검사 부적합률의 지역적 영향요인 분석
김우석(Woosuk Kim),김도경(Do-Gyeong Kim),박정수(Jungsoo Park) 한국자동차공학회 2022 한국 자동차공학회논문집 Vol.30 No.1
A type of spatial dependence might be suspected in the vehicle inspection data because it has similar characteristics with spatial data. This study aims to contribute to the establishment of a traffic operation order by revealing the spatial autocorrelation and by identifying regional characteristics that influence vehicle inspection failure rates. Based on the estimation of spatial econometric models, spatial dependence was found with a value of 0.37(Moran’s I index), indicating that vehicle inspection data are spatially correlated. With respect to regional characteristics affecting vehicle inspection failure rates, five significant factors were identified: average vehicle age, average temperature, percentage of private inspection stations, percentage of diesel vehicles, and amount of precipitation. The results showed that differentiated vehicle inspections according to the characteristics of each region, such as strengthening automobile fuel filter inspections in areas with low average temperatures and strengthening emission inspections in regions with a high proportion of diesel vehicles, should be conducted.
고해상도 영상 압축을 위한 SPIHT 기반의 부대역 분할 압축 방법
김우석(Woosuk Kim),박병서(Byung-Seo Park),오관정(Kwan-Jung Oh),서영호(Young-Ho Seo) 한국방송·미디어공학회 2022 방송공학회논문지 Vol.27 No.2
This paper proposes a method to solve problems that may occur when SPIHT(set partition in hierarchical trees) is used in a dedicated codec for compressing complex holograms with ultra-high resolution. The development of codecs for complex holograms can be largely divided into a method of creating dedicated compression methods and a method of using anchor codecs such as HEVC and JPEG2000 and adding post-processing techniques. In the case of creating a dedicated compression method, a separate conversion tool is required to analyze the spatial characteristics of complex holograms. Zero-tree-based algorithms in subband units such as EZW and SPIHT have a problem that when coding for high-resolution images, intact subband information is not properly transmitted during bitstream control. This paper proposes a method of dividing wavelet subbands to solve such a problem. By compressing each divided subbands, information throughout the subbands is kept uniform. The proposed method showed better restoration results than PSNR compared to the existing method.
김우석(Woosuk Kim),권승준(Kwon, Seung Joon) 한국정보기술학회 2021 Proceedings of KIIT Conference Vol.2021 No.11
핸드 제스처 인식은 가상 현실 콘텐츠에서 자연스러운 상호작용을 가능하게 하는 주요 기술 중 하나이다. 본 논문에서는 스켈레톤 형태로 주어지는 제스처의 인식을 위한 심층 신경망 모델을 제시한다. 그래프 형태로 기술되는 핸드 스켈레톤 입력에 대해 특징값을 추출하기 위한 GNN기반의 구조를 기술하고 이를 통해 계산된 특징값 시퀀스로부터 제스처 종류를 구분하기 위한 모델을 구성하였다. 설계된 모델의 인식 성능은 공개되어 있는 두가지 데이터 셋인 DHG-14/28 및 SHREC’17에 대해 학습한 결과를 토대로 확인하였다. Hand gesture recognition is one of the important techniques for natural and expressive interaction in virtual environments. In this paper, we propose a method to recognize gestures given in the form of skeletons. A GNN based model which calculates features of hand postures from skeletal input represented as graphs is presented and a classification model which discriminates gesture types from the feature sequence is described. The recognition performance was measured using two public dataset, which are DHG-14/28 and SHREC’17.