• 한국지구물리탐사학회:학술대회논문집
• /
• 2007.06a
• /
• pp.54-57
• /
• 2007

Recently, multidimensional joint inversion of geophysical data based on fundamentally different physical properties has been actively studied. Joint inversion can provide a way to much more accurately image the subsurface structure. Through the joint inversion, furthermore, it is possible to directly estimate non-geophysical material properties from geophysical measurements. In this study, I derive the objective functions and normal equations of three different joint inversion approaches: one approach based on the structural similarity using cross-gradient, and the other two using the a priori information on the model parameters and the correlation between material properties. Since all the equations derived in this study are based on the same inversion method (smoothness constrained least-squares), it is possible to mix the joint inversion methods so as to produce a new joint inversion algorithm.

• 한국지구물리탐사학회:학술대회논문집
• /
• 2007.06a
• /
• pp.124-129
• /
• 2007

For seismic imaging and inversion, the inverted image depends on how we define the objective function. ${\ell}^1$-norm is more robust than ${\ell}^2$-norm. However, it is difficult to apply the Newton-type algorithm directly because the partial derivative for ${\ell^1$-norm has a singularity. In our paper, to overcome the difficulties of singularities, Huber function given by hybrid ${\ell}^1/{\ell}^2$-norm is used. We tested the robustness of our new object function with several noisy data set. Numerical results show that the new objective function is more robust to band limited spiky noise than the conventional object function.

• Proceedings of the Korean Nuclear Society Conference
• /
• 1997.10a
• /
• pp.329-334
• /
• 1997

A tomography algorithm to maximize the entropy of image using Lagrangian multiplier technique and conjugate gradient method has been designed for the measurement of 2D spatial distribution of intense neutral beams of KSTAR NBI(Korea Superconducting Tokamak Advanced Research Neutral Beam Injector) which is now being designed. A possible detection system was assumed and a numerical simulation has been implemented to test the reconstruction quality of given beam profiles. This algorithm has the good applicability for sparse projection data and thus, can be used for the neutral beam tomography.

• International Journal of Control, Automation, and Systems
• /
• v.4 no.5
• /
• pp.660-667
• /
• 2006

This paper presents a novel region-based snake method for vessel segmentation. According to geometric shape analysis of the vessel structure with different scale, an efficient statistical estimation of vessel branches is introduced into the energy objective function, which applies not only the vessel intensity information, but also geometric information of line-like structure in the image. The defined energy function is minimized using the gradient descent method and a new region-based speed function is obtained, which is more accurate to the vessel structure and not sensitive to the initial condition. The narrow band algorithm in the level set framework implements the proposed method, the solution of which is steady. The segmentation experiments are shown on several images. Compared with other geometric active contour models, the proposed method is more efficient and robust.

• Journal of the Korea Computer Graphics Society
• /
• v.11 no.1
• /
• pp.40-49
• /
• 2005

Several optimization techniques have been proposed for volume ray casting, but these cannot achieve real-time frame rates. In addition, it is difficult to apply them to some applications that require perspective projection. Recently, hardware-based methods using 3D texture mapping are being used for real-time volume rendering. Although rendering speed approaches real time, the larger volumes require more swapping of volume bricks for the limited texture memory. Also, image quality deteriorates compared with that of conventional volume ray casting. In this paper, we propose a data structure for real-time volume ray casting named PERM (Precomputed dEnsity and gRadient Map). The PERM stores interpolated density and gradient vector for quantized cells. Since the information requiring time-consuming computations is stored in the PERM, our method can ensure interactive frame rates on a consumer PC platform. Our method normally produces high-quality images because it is based on conventional volume ray casting.

• Proceedings of the KIEE Conference
• /
• 2007.10a
• /
• pp.117-118
• /
• 2007

A robot needs a human detection algorithm for interaction with a human. This paper proposes a method that finds people using a SVM (support vector machine) classifier and a stereo camera. Feature vectors of SVM are extracted by HoG (histogram of gradient) within images. After training extracted vectors from the clustered images, the SVM algorithm creates a classifier for human detection. Each candidate for a human in the image is generated by clustering of depth information from a stereo camera and the candidate is evaluated by the classifier. When compared with the existing method of creating candidates for a human, clustering reduces computational time. The experimental results demonstrate that the proposed approach can be executed in real time.

• Proceedings of the KIEE Conference
• /
• 1998.11b
• /
• pp.597-599
• /
• 1998

The clinical acceptance of magnetic resonance imaging(MRI) system has been more rapid than that of the other medical image diagnosis system(X-ray, CT, etc) with the advantage of nonhazardous nature, high resolution capability, potential for chemically specified imaging. MRI system is composed of super conducting magnet, gradient fields, rf transceiver, system controller and imaging software technology. In this paper, introducing the principle of magnetic resonance imaging, it proposes the implementation of PC-based MRI system.

• Proceedings of the Korean Institute of Intelligent Systems Conference
• /
• 1998.10a
• /
• pp.468-473
• /
• 1998

An intelligent multimode target tracking algorithm using fuzzy logic is presented. Multimode tracking represents a synergistic approach that utilizes a variety of tracking techniques(centroid, correlation, etc.) to overcome the limitations inherent in any single-mode tracker. The design challenge for this type of multimode tracker is the data fusion algorithm. designs for this algorithm are based on heuristic rather than analytical approaches. A correlation-tracking algorithm seeks to align the incoming target image with a reference in age of the target, but has a critical problem, so called drift phenomenon. In this paper we will suggest a robust correlation tracker with gradient preprocessor combined by centroid algorithm to overcome the drift problem.

• Annual Conference of KIPS
• /
• 2004.11a
• /
• pp.847-850
• /
• 2004

본 논문에서는 모폴로지(Morphology) 재구성(Reconstruction)과 비선형 확산(Non-Linear Diffusion)을 이용하여 칼라 영상을 유사한 영역으로 분할하는 방법을 제안한다. 초기에 RGB 영상을 LUV 색상 공간으로 전환하고, 그 색상공간에 모폴로지를 응용한 재구성(Reconstruction)에 의한 닫힘(Closing) 연산과 비선형 확산(Non-Linear Diffusion)을 적용하여 잡음을 제거한 실험 영상을 획득한다. 이 영상에서 워터쉐드 알고리즘을 위한 칼라 영상의 기울기(Gradient) 정보를 획득하고, 그 영상에 마커(Marker) 정보를 이용한 워터쉐드(Watershed) 알고리즘을 적용하여 영상을 효과적으로 분할한다. 칼라 영상을 대상으로 한 실험에서 제안 방법이 영상을 효과적으로 분할함을 확인 하였다.

• Annual Conference of KIPS
• /
• 2019.10a
• /
• pp.795-797
• /
• 2019

In this paper, we propose a method to generate adversarial examples for toxicity detection neural networks. Our dataset is represented by a one-hot vector and we constrain that only one character is allowed to be modified. The location to be changed is founded by the maximum area of input gradient, which represents the most affecting character the model to make decisions. Despite the fact that we have strong constraint compared to the image-based adversarial attack, we have achieved about 49% successful rate.