• KIPS Transactions on Software and Data Engineering
• /
• v.13 no.1
• /
• pp.17-27
• /
• 2024

Video frame interpolation is an important technique used in the field of video and media, as it increases the continuity of motion and enables smooth playback of videos. In the study of video frame interpolation using deep learning, Kernel Based Method captures local changes well, but has limitations in handling global changes. In this paper, we propose a new U-Net structure that applies feature map differentiation and two directions to focus on capturing major changes to generate intermediate frames more accurately while reducing the number of parameters. Experimental results show that the proposed structure outperforms the existing model by up to 0.3 in PSNR with about 61% fewer parameters on common datasets such as Vimeo, Middle-burry, and a new YouTube dataset. Code is available at https://github.com/Go-MinSeong/SF-AdaCoF.

• Journal of Broadcast Engineering
• /
• v.23 no.4
• /
• pp.559-573
• /
• 2018

Recently, one of the most actively applied image media in the most fields such as virtual reality (VR) is omni-directional or panorama image. This image is generated by stitching images obtained by various methods. In this process, it takes the most time to extract keypoints necessary for stitching. In this paper, we analyze the parameters involved in the extraction of SIFT keypoints with the aim of reducing the computation time for extracting the most widely used SIFT keypoints. The parameters considered in this paper are the initial standard deviation of the Gaussian kernel used for Gaussian filtering, the number of gaussian difference image sets for extracting local extrema, and the number of octaves. As the SIFT algorithm, the Lowe scheme, the originally proposed one, and the Hess scheme which is a convolution cascade scheme, are considered. First, the effect of each parameter value on the computation time is analyzed, and the effect of each parameter on the stitching performance is analyzed by performing actual stitching experiments. Finally, based on the results of the two analyses, we extract parameter value set that minimize computation time without degrading.

• The Journal of Korea Institute of Information, Electronics, and Communication Technology
• /
• v.15 no.5
• /
• pp.380-386
• /
• 2022

This paper proposes a method for structuring the preprocessing process of a training image when color is applied using Pix2Pix, one of the adversarial generative neural network techniques. This paper concentrate on the prediction result can be damaged according to the degree of light reflection of the training image. Therefore, image preprocesisng and parameters for model optimization were configured before model application. In order to increase the image resolution of training and prediction results, it is necessary to modify the of the model so this part is designed to be tuned with parameters. In addition, in this paper, the logic that processes only the part where the prediction result is damaged by light reflection is configured together, and the pre-processing logic that does not distort the prediction result is also configured.Therefore, in order to improve the usability, the accuracy was improved through experiments on the part that applies the light reflection tuning filter to the training image of the Pix2Pix model and the parameter configuration.

• Journal of the Computational Structural Engineering Institute of Korea
• /
• v.32 no.1
• /
• pp.55-63
• /
• 2019

In engineering problems, many random variables have correlation, and the correlation of input random variables has a great influence on reliability analysis results of the mechanical systems. However, correlated variables are often treated as independent variables or modeled by specific parametric joint distributions due to difficulty in modeling joint distributions. Especially, when there are insufficient correlated data, it becomes more difficult to correctly model the joint distribution. In this study, multivariate kernel density estimation with bounded data is proposed to estimate various types of joint distributions with highly nonlinearity. Since it combines given data with bounded data, which are generated from confidence intervals of uniform distribution parameters for given data, it is less sensitive to data quality and number of data. Thus, it yields conservative statistical modeling and reliability analysis results, and its performance is verified through statistical simulation and engineering examples.

• Journal of the Korea Institute of Information and Communication Engineering
• /
• v.17 no.10
• /
• pp.2425-2430
• /
• 2013

This paper presents an shifted linear interpolation method with an image-dependent parameter. The previous shifted linear interpolation proposed the optimal shift parameter of 0.21, which is calculated by spectrum analysis of the shifted linear interpolation kernel. However, the parameter can be different if we takes an input image spectrum into account. Thus, we introduce an image-dependent parameter. An experiment shows the best shift parameter is 0.19 in average for real images. Also, simulation results indicate the proposed method is superior to the existing shifted linear interpolation as well as conventional methods such as linear interpolation and cubic convolution interpolation in terms of the subjective and objective image quality.

• Journal of IKEEE
• /
• v.24 no.4
• /
• pp.1022-1027
• /
• 2020

In this paper, SVM-based speaker classification is experimented with GMM-supervector. To create a speaker cluster, conventional speaker change detection is performed with the KL distance using the SNR-based weighting function. SVM-based speaker classification consists of two steps. In the first step, SVM-based classification between UBM and speaker models is performed, speaker information is indexed in each cluster, and then grouped by speaker. In the second step, the SVM-based classification between UBM and speaker models is performed by inputting the speaker cluster group. Linear and RBF are applied as kernel functions for SVM-based classification. As a result, in the first step, the case of applying the linear kernel showed better performance than RBF with 148 speaker clusters, MDR 0, FAR 47.3, and ER 50.7. The second step experiment result also showed the best performance with 109 speaker clusters, MDR 1.3, FAR 28.4, and ER 32.1 when the linear kernel was applied.

• Proceedings of the Korean Society of Broadcast Engineers Conference
• /
• 2010.11a
• /
• pp.302-303
• /
• 2010

본 논문에서는 고속으로 홀로그램을 생성하기 위해 새로운 컴퓨터 생성 홀로그램(computer-generated hologram, CGH) 수식을 제안하고, 셀 기반의 VLSI(very large scale integrated circuit) 구조를 제안하였다. 기본 CGH 수식에서 가로 또는 세로 방향의 연산 규칙을 찾아낸 후 가로 또는 세로 방향의 홀로그램 화소를 병렬적으로 구할 수 있는 수식을 유도하였다. 제안한 수식을 바탕으로 초기 파라미터 연산기(initial parameter calculator)와 업데이트-위상 연산기(update-phase calculator)로 구성된 CGH 셀의 구조를 제안하고 하드웨어로 구현하였다. 수식의 변형을 통해서 하드웨어를 간략화 시킬 수 있었고, CGH의 확장을 통해 가로 방향으로 병렬화시킬 수 있는 하드웨어 구조도 보였다. 실험에서는 하드웨어에 사용된 자원을 분석하였다. CGH 커널과 프로세서의 구조는 이전 연구에서 사용된 플랫폼을 그대로 사용하였다.

• Proceedings of the IEEK Conference
• /
• 2009.05a
• /
• pp.261-263
• /
• 2009

본 연구에서는 복잡한 비선형 모델링 방법인 RBF 뉴럴 네트워크(Radial Basis Function Neural Network)와 PNN(Polynomial Neural Network)을 접목한 새로운 형태의 Radial Basis Function Polynomial Neural Network(RPNN)를 제안한다. RBF 뉴럴 네트워크는 빠른 학습 시간, 일반화 그리고 단순화의 특징으로 비선형 시스템 모델링 등에 적용되고 있으며, PNN은 생성된 노드들 중에서 우수한 결과값을 가진 노드들을 선택함으로써 모델의 근사화 및 일반화에 탁월한 효과를 가진 비선형 모델링 방법이다. 제안된 RPNN모델의 기본적인 구조는 PNN의 형태를 이루고 있으며, 각각의 노드는 RBF 뉴럴 네트워크로 구성하였다. 사용된 RBF 뉴럴 네트워크에서의 커널 함수로는 FCM 클러스터링을 사용하였으며, 각 노드의 후반부는 다항식 구조로 표현하였다. 또한 각 노드의 후반부 파라미터들은 최소자승법을 이용하여 최적화 하였다. 제안한 모델의 적용 및 유용성을 비교 평가하기 위하여 비선형 데이터를 이용하여 그 우수성을 보인다.

• Journal of the Korea Safety Management & Science
• /
• v.11 no.1
• /
• pp.87-91
• /
• 2009

본 논문은 상대적으로 새로운 기법인 Parzen Density Estimation과 Multi-class SVM을 이용한 지능형 고장 탐색과 진단 방법을 제안하고 있다. 본 연구에서는 롤링 베어링을 대상으로 고장을 탐색하고 진단하기 위한 방법을 제안하는데 Parzen Density Estimation과 Multi-class SVM은 고장 클래스를 잘 표현할 수 있다. Parzen Density Estimation은 새로운 패턴 데이터의 거절과 알려진 데이터 패턴의 밀도의 평가에 의해 새로운 패턴을 찾아낼 수 있고, Multi-class SVM 기반의 방법은 여러 클래스의 고장을 support vector로 표현하여 고장 패턴을 찾아낼 수 있다. 본 연구에서는 실제의 다중 클래스를 가지는 롤링 베어링의 고장 데이터를 사용하여 고장 패턴을 탐색하는 과정을 보여주는데, 커널함수의 적절한 파라미터의 선택에 의한 Multi-class SVM 기반의 방법이 multi-layer perceptron이나 Parzen Density Estimation 방법보다 우수함을 입증한다.

• Annual Conference of KIPS
• /
• 2011.04a
• /
• pp.967-970
• /
• 2011

본 논문에서는 커널 모드에서 악성 봇이 호스트를 전염 시키는 순간 나타나는 일반적인 행동 특성들을 기반으로 효과적인 악성 봇 탐지가 가능한 프로그램을 구현하였다. 구현된 프로그램은 false-positive(오탐지)를 줄이기 위해서 악성 봇의 전염 과정에서 발생하는 복제 행동, 레지스트리 등록, uninstall 등록, 복제된 파일의 경로 정보 그리고 사용할 API 임포트 정보 등과 같은 악성 행위 탐지 기준 6가지를 고려한다.