• 디지털산업정보학회논문지
• /
• 제4권1호
• /
• pp.25-31
• /
• 2008

In this paper, we propose a new method that separates background and foreground from stereo images. This method can be improved automatic target tracking system by using disparity map of the stereo vision system and background-separating mask, which can be obtained camera configuration parameters. We use disparity map and camera configuration parameters to separate object from background. Disparity map is made with block matching algorithm from stereo images. A morphology filter is used to compensate disparity error that can be caused by occlusion area. We could obtain a separated object from background when the proposed method was applied to real stereo cameras system.

• 한국CDE학회논문집
• /
• 제14권1호
• /
• pp.10-17
• /
• 2009

Surface detection and parameter estimation in point cloud is a relevant subject in CAD/CAM, reverse engineering, computer vision, coordinate metrology and digital factory. In this paper we present a software for a fully automatic surface detection and parameter estimation in unordered, incomplete and error-contaminated point cloud with a large number of data points. The software consists of three algorithmic modules each for object identification, point segmentation, and model fitting, which work interactively. Our newly developed algorithms for orthogonal distance fitting(ODF) play a fundamental role in each of the three modules. The ODF algorithms estimate the model parameters by minimizing the square sum of the shortest distances between the model feature and the measurement points. We demonstrate the performance of the software on a variety of point clouds generated by laser radar, computer tomography, and stripe-projection method.

• Journal of Multimedia Information System
• /
• 제8권1호
• /
• pp.31-34
• /
• 2021

The classical HMM is defined by a parameter triple �� = (��, A, B), where each parameter represents a collection of probability distributions: initial state, state transition and output distributions in order. This paper proposes a new stationary parameter e = (e1, e2, …, eN) where N is the number of states and et = P(|xt = i, y) for describing how an input pattern y ends in state xt = i at time t followed by nothing. It is often said that all is well that ends well. We argue here that all should end well. The paper sets the framework for the theory and presents an efficient inference and training algorithms based on dynamic programming and expectation-maximization. The proposed model is applicable to analyzing any sequential data with two or more finite segmental patterns are concatenated, each forming a context to its neighbors. Experiments on online Hangul handwriting characters have proven the effect of the proposed augmentation in terms of highly intuitive segmentation as well as recognition performance and 13.2% error rate reduction.

• 한국멀티미디어학회논문지
• /
• 제25권4호
• /
• pp.568-574
• /
• 2022

In this paper, direction normalization and CNN deep learning are used to develop a more reliable license plate recognition system. The existing license plate recognition system consists of three main modules: license plate detection module, character segmentation module, and character recognition module. The proposed system minimizes recognition error by adding a direction normalization module when a detected license plate is inclined. Experimental results show the superiority of the proposed method in comparison to the previous system.

• 대한전자공학회논문지
• /
• 제26권2호
• /
• pp.87-93
• /
• 1989

패턴인식과 영상처리 응용에 이용되는 자승오차 패턴 클러스터링 알고리듬은 특징벡터 행렬의 연산에 상당한 처리시간은 요구한다. 그러므로 본 논문은 병렬처리와 파이프라인 특성을 갖는 잉여수체계를 이용한 고속의 자승오차 패턴 클러스터링 프로세서를 제안한다. 제안된 자승오차 패턴 클러스터링 프로세서는 영상분할 실험으로부터 의미있는 영역으로 나눌 수 있는 클러스터의 수에 대하여 만족할 만한 오차를 보이며 80287 수치 연산용 프로세서보다 약 200배 빠름을 보인다. 그 결과 대규모의 데이타를 실시간으로 처리하여야 하는 응용분야에 효과적으로 이용할 수 있음을 확인하였다.

• Korean Journal of Radiology
• /
• 제21권1호
• /
• pp.88-100
• /
• 2020

Objective: We aimed to develop and validate a deep learning system for fully automated segmentation of abdominal muscle and fat areas on computed tomography (CT) images. Materials and Methods: A fully convolutional network-based segmentation system was developed using a training dataset of 883 CT scans from 467 subjects. Axial CT images obtained at the inferior endplate level of the 3rd lumbar vertebra were used for the analysis. Manually drawn segmentation maps of the skeletal muscle, visceral fat, and subcutaneous fat were created to serve as ground truth data. The performance of the fully convolutional network-based segmentation system was evaluated using the Dice similarity coefficient and cross-sectional area error, for both a separate internal validation dataset (426 CT scans from 308 subjects) and an external validation dataset (171 CT scans from 171 subjects from two outside hospitals). Results: The mean Dice similarity coefficients for muscle, subcutaneous fat, and visceral fat were high for both the internal (0.96, 0.97, and 0.97, respectively) and external (0.97, 0.97, and 0.97, respectively) validation datasets, while the mean cross-sectional area errors for muscle, subcutaneous fat, and visceral fat were low for both internal (2.1%, 3.8%, and 1.8%, respectively) and external (2.7%, 4.6%, and 2.3%, respectively) validation datasets. Conclusion: The fully convolutional network-based segmentation system exhibited high performance and accuracy in the automatic segmentation of abdominal muscle and fat on CT images.

• 방송공학회논문지
• /
• 제22권5호
• /
• pp.608-617
• /
• 2017

영역 분리에 의한 명암대비 방법들이 제안되어 왔지만 영상의 히스토그램에 따라 과포화 되는 부작용이나 밝기 값 보존과 명암대비 효과의 상반 관계에 대한 개선이 필요하다. 본 논문은 다양한 히스토그램에서도 명암 대비가 개선 되도록 영역 분리 시 각 서브 영역이 가우시안 분포를 갖도록 분리하고 영역별 평활화하는 명암 대비 방법을 제안 한다. 영역 분리는 $L^*a^*b^*$ 컬러 공간에서 K-평균 방법과 기대-최대 방법에 의해 영역맵과 확률맵을 생성하며 영역별 히스토그램 평활화 방법은 영역간 히스토그램 중복 최소를 위해 평균값 이동과 영역 분리에서 생성된 확률맵을 변환 함수에 활용함으로써 영역별 밝기값을 보존 하였다. 실험은 기존의 명암 대비 방법들과 평균 밝기 차이와 평균 엔트로피 값을 이용하여 밝기 변화가 적고 영상의 세부 정보가 표현됨에 의한 명암대비 개선을 보인다.

• 말소리와 음성과학
• /
• 제8권1호
• /
• pp.71-80
• /
• 2016

The phonemic awareness tasks (phonemic synthesis, phonemic elision, phonemic segmentation) by auditory presentation and visual presentation were conducted to 40 children who are 5 and 6 years old. The scores and error types in the sub-tasks by two presentations were compared to each other. Also, the correlation between the performances of phonemic awareness sub-tasks in two presentation conditions were examined. As a result, 6-year-old group showed significantly higher phonemic awareness scores than 5-year-old group. Both group showed significantly higher scores in visual presentation than auditory presentation. While the performance under the visual presentation was significantly lower especially in the segmentation than the other two tasks, there was no significant difference among sub-tasks under the auditory presentation. 5-year-old group showed significantly more 'no response' errors than 6-year-old group and 6-year-old group showed significantly more 'phoneme substitution' and 'phoneme omission' errors than 5-year-old group. Significantly more 'phoneme omission' errors were observed in the segmentation than the elision task, and significantly more 'phoneme addition' errors were observed in elision than the synthesis task. Lastly, there are positive correlations in auditory and visual synthesis tasks, auditory and visual elision tasks, and auditory and visual segmentation tasks. Summarizing the results, children tend to depend on orthographic knowledge when acquiring the initial phonemic awareness. Therefore, the result of this research would support the position that the orthographic knowledge affects the improvement of phonemic awareness.

• 대한의용생체공학회:의공학회지
• /
• 제40권5호
• /
• pp.179-188
• /
• 2019

In ultrasound telemedicine, it is important to reduce the size of the data by compressing the ultrasound image when sending it. Ultrasound images can be divided into image context and other information consisting of patient ID, date, and several letters. Between them, ultrasound context is very important information for diagnosis and should be securely preserved as much as possible. In several previous papers, ultrasound compression methods were proposed to compress ultrasound context and other information into different compression parameters. This ultrasound compression method minimized the loss of ultrasound context while greatly compressing other information. This paper proposed the method of automatic segmentation of ultrasound context to overcome the limitation of the previously described ultrasound compression method. This algorithm was designed to robust for various ultrasound device and to enable real-time operation to maintain the benefits of ultrasound imaging machine. The operation time of extracting ultrasound context through the proposed segmentation method was measured, and it took 311.11 ms. In order to optimize the algorithm, the ultrasound context was segmented with down sampled input image. When the resolution of the input image was reduced by half, the computational time was 126.84 ms. When the resolution was reduced by one-third, it took 45.83 ms to segment the ultrasound context. As a result, we verified through experiments that the proposed method works in real time.

• 한국정보통신학회:학술대회논문집
• /
• 한국정보통신학회 2021년도 춘계학술대회
• /
• pp.474-476
• /
• 2021

신경 교세포에서 발생하는 가장 흔한 뇌 악성종양인 교모세포종은 조기 진단 및 치료계획 수립이 중요하다. 주로 조영제를 통해 촬영된 T1CE 영상을 통해 암의 진단이 이뤄지는데, 최근 가돌리늄 기반 조영제 주입의 위험이 보고되고 있다. 의료영상에서 새로운 영상을 합성하는 GAN 모델과 영역분할에 대한 심층신경망 모델에 대한 연구가 활발히 진행되고 있다. 본 연구에서는 교모세포종 환자의 T1CE 영상의 생성과 암의 영역분할을 동시에 학습하는 하나의 모델을 제안한다. 제안된 모델의 성능은 평균 제곱오차, 최대신호대잡음비 등의 유사성 측정을 통해 평가되어 0.002, 55dB의 평균 결과 값을 보여준다.