• The Journal of Korean Institute of Communications and Information Sciences
• /
• v.22 no.9
• /
• pp.1985-1997
• /
• 1997

This paper proposed a regularized iterative image restoration by using method of conjugate gradient. Compared with conventional iterative methods, method of conjugate gradient has a merit to converte toward a solution as a super-linear convergence speed. But because of those properties, there are several artifacts like ringing effects and the partial magnification of the noise in the course of restoring the images that are degraded by a defocusing blur and additive noise. So, we proposed the regularized method of conjugate gradient applying constraints. By applying the projectiong constraint and regularization parameter into that method, it is possible to suppress the magnification of the additive noise. As a experimental results, we showed the superior convergence ratio of the proposed mehtod compared with conventional iterative regularized methods.

• The Journal of Korean Institute of Communications and Information Sciences
• /
• v.21 no.10
• /
• pp.2541-2551
• /
• 1996

Human visual systems are sensitive to noise on the flat intensity area. But it becomes less sensitive on the edge area. Recently, many types of spatially adaptive image restoration methods have been proposed, which employ the above mentioned huan visual characteristics. The present paper presents an adaptive image restoration method, which increases sharpness of the edge region, and smooths noise on the flat intensity area. For edge detection, the proposed method uses the visibility function based on the local variance on each pixel. And it adaptively changes the regularization parameter. More specifically, the image to be restored is divided into a number of steps from the flat area to the edge regio, and then restored by using the finite impulse response constrained least squares filter.

• Journal of Astronomy and Space Sciences
• /
• v.30 no.4
• /
• pp.321-326
• /
• 2013

The modulation transfer function (MTF) is a widely used indicator in assessments of remote-sensing image quality. This MTF method is also used to restore information to a standard value to compensate for image degradation caused by atmospheric or satellite jitter effects. In this study, we evaluated MTF values as an image quality indicator for the Geostationary Ocean Color Imager (GOCI). GOCI was launched in 2010 to monitor the ocean and coastal areas of the Korean peninsula. We evaluated in-orbit MTF value based on the GOCI image having a 500-m spatial resolution in the first time. The pulse method was selected to estimate a point spread function (PSF) with an optimal natural target such as a Seamangeum Seawall. Finally, image restoration was performed with a Wiener filter (WF) to calculate the PSF value required for the optimal regularization parameter. After application of the WF to the target image, MTF value is improved 35.06%, and the compensated image shows more sharpness comparing with the original image.

• Structural Engineering and Mechanics
• /
• v.81 no.1
• /
• pp.103-115
• /
• 2022

The prediction error variances for frequencies are usually considered as unknown in the Bayesian system identification process. However, the error variances for mode shapes are taken as known to reduce the dimension of an identification problem. The present study attempts to explore the effectiveness of Bayesian approach of model parameters updating using Markov Chain Monte Carlo (MCMC) technique considering the prediction error variances for both the frequencies and mode shapes. To remove the ergodicity of Markov Chain, the posterior distribution is obtained by Gaussian Random walk over the proposal distribution. The prior distributions of prediction error variances of modal evidences are implemented through inverse gamma distribution to assess the effectiveness of estimation of posterior values of model parameters. The issue of incomplete data that makes the problem ill-conditioned and the associated singularity problem is prudently dealt in by adopting a regularization technique. The proposed approach is demonstrated numerically by considering an eight-storey frame model with both complete and incomplete modal data sets. Further, to study the effectiveness of the proposed approach, a comparative study with regard to accuracy and computational efficacy of the proposed approach is made with the Sequential Monte Carlo approach of model parameter updating.

• Proceedings of the Korean Institute of Information and Commucation Sciences Conference
• /
• 2022.10a
• /
• pp.127-130
• /
• 2022

This paper is a study on singing voice synthesis modeling using a generator loss function, which analyzes various factors that may occur when applying BEGAN among deep learning algorithms optimized for image generation to Audio domain. and we conduct experiments to derive optimal quality. In this paper, we focused the problem that the L1 loss proposed in the BEGAN-based models degrades the meaning of hyperparameter the gamma(𝛾) which was defined to control the diversity and quality of generated audio samples. In experiments we show that our proposed method and finding the optimal values through tuning, it can contribute to the improvement of the quality of the singing synthesis product.

• Journal of the Korean earth science society
• /
• v.30 no.7
• /
• pp.834-844
• /
• 2009

Two-dimensional magnetotelluric (MT) inversion algorithm using minimum support (MS) stabilizer functional was implemented in this study to enhance the contrast of inverted images. For this implementation, this study derived a formula in discrete form for creeping model updates in the least-squares linearized inversion. A spatially varying regularization parameter determination algorithm, which is known as ACB (Active Constraint Balancing), was also adopted to stabilize the inversion process when using MS stabilizer as a model constraint. Inversion experiments for a simple isolated body model show well the feature of MS stabilizer in concentrating the anomalous body compared with the second-order derivative model constraint. This study also compared MS stabilizer and the second-order derivative model constraints for a model having multiple anomalous bodies to show the applicability of the algorithm into field data.

• Geophysics and Geophysical Exploration
• /
• v.8 no.2
• /
• pp.129-136
• /
• 2005

The damped least-squares inversion has become a most popular method in finding the solution in geophysical problems. Generally, the least-squares inversion is to minimize the object function which consists of data misfits and model constraints. Although both the data misfit and the model constraint take an important part in the least-squares inversion, most of the studies are concentrated on what kind of model constraint is imposed and how to select an optimum regularization parameter. Despite that each datum is recommended to be weighted according to its uncertainty or error in the data acquisition, the uncertainty is usually not available. Thus, the data weighting matrix is inevitably regarded as the identity matrix in the inversion. We present a new inversion scheme, in which the data weighting matrix is automatically obtained from the analysis of the data resolution matrix and its spread function. This approach, named 'extended active constraint balancing (EACB)', assigns a great weighting on the datum having a high resolution and vice versa. We demonstrate that by applying EACB to a two-dimensional resistivity tomography problem, the EACB approach helps to enhance both the resolution and the stability of the inversion process.

• Progress in Superconductivity and Cryogenics
• /
• v.22 no.1
• /
• pp.12-16
• /
• 2020

As high homogeneity in magnetic field is required to increase the resolution of MRI magnets, various shimming methods have been researched. Using one of them, the design of the superconducting active zonal shim coil for MRI magnets is discussed in this paper. The magnetic field of the MRI magnet is expressed as the sum of spherical harmonic terms, and the optimized current density of shim coils capable of removing higher-order terms is calculated by the Tikhonov regularization method. To investigate all potential designs derived from calculated current density, 4 sweeping parameters are selected: (1) axial length of shim coil zone; (2) radius of shim coils; (3) exact axial position of shim coils; and (4) operating current. After adequate designs are determined with constraints of critical current margin and homogeneity criterion, the total wire length required for each is calculated and the design with a minimum of them is chosen. Using the superconducting wire length of 9.77 km, the field homogeneity over 50 cm DSV is improved from 24 ppm to 1.87 ppm in the case study for 9.4 T whole-body MRI shimming. Finally, the results are compared with the finite element method (FEM) simulation results to validate the feasibility and accuracy of the design.

• Journal of the Korean Data and Information Science Society
• /
• v.12 no.2
• /
• pp.1-10
• /
• 2001

Neural networks we increasingly being seen as an addition to the statistics toolkit which should be considered alongside both classical and modern statistical methods. Neural networks are usually useful for classification and function estimation. In this paper we concentrate on function estimation using neural networks with weight decay factor The use of weight decay seems both to help the optimization process and to avoid overfitting. In this type of neural networks, the problem to decide the number of hidden nodes, weight decay parameter and iteration number of learning is very important. It is called the optimization of weight decay neural networks. In this paper we propose a automatic optimization based on genetic algorithms. Moreover, we compare the weight decay neural network automatically learned according to automatic optimization with ordinary neural network, projection pursuit regression and support vector machines.

• Proceedings of the Korean Institute of Information and Commucation Sciences Conference
• /
• 2018.10a
• /
• pp.626-628
• /
• 2018

In this paper, we compare the performances of SVM (Support Vector Machine) and ANN (Artificial Neural Network) machine learning models for predicting solar energy by using meteorological data. Two machine learning models were built by using fifteen kinds of weather data such as long and short wave radiation average, precipitation and temperature. Then the RBF (Radial Basis Function) parameters in the SVM model and the number of hidden layers/nodes and the regularization parameter in the ANN model were found by experimental studies. MAPE (Mean Absolute Percentage Error) and MAE (Mean Absolute Error) were considered as metrics for evaluating the performances of the SVM and ANN models. Sjoem Simulation results showed that the SVM model achieved the performances of MAPE=21.11 and MAE=2281417.65, and the ANN model did the performances of MAPE=19.54 and MAE=2155345.10776.