• Proceedings of the Korean Society of Broadcast Engineers Conference
• /
• 2021.06a
• /
• pp.176-179
• /
• 2021

최근 딥 러닝 기법의 하나인 합성곱 신경망(Convolutional Neural Network, CNN)은 영상 잡음(Noise) 제거 분야에서 전통적인 기법보다 좋은 성능을 나타내고 있지만 학습하는 과정에서 영상 내 디테일한 부분이 손실될 수 있다. 본 논문에서는 웨이블릿 변환(Wavelet Transform)을 기반으로 영상 내 디테일 정보도 같이 학습하여 영상 디테일을 향상하는 잡음 제거 합성곱 신경망 네트워크를 제안한다. 제안하는 네트워크는 디테일 향상 서브 네트워크(Detail Enhancement Subnetwork)와 영상 잡음 추출 서브 네트워크(Noise Extraction Subnetwork)를 이용하게 된다. 실험을 통해 제안하는 방법은 기존 알고리듬보다 디테일 손실 문제를 효과적으로 해결할 수 있었고 객관적 품질 평가인 PSNR(Peak Signal-to-Noise Ratio)와 주관적 품질 비교에서 모두 우수한 결과가 나온 것을 확인하였다.

• Journal of the Korean Society of Surveying, Geodesy, Photogrammetry and Cartography
• /
• v.9 no.2
• /
• pp.103-113
• /
• 1991

The accuracy of input coordinates of ground control points and check points affects great influences to the results of ground coordinate computation in using SPOT digital image data. The original SPOT images displayed on CRT are not usually adequate for identifying the object features and determining the point positioning. Hence, appropriate image processing techniques such as contrast enhancement, subpixel interpolation, edge enhancement, and spatial filtering are needed. In this study, the principles of digital image processing needed for accurate three dimensional positioning and spectral characteristic analysis are investigated. The algorithms for the actual applications are developed and programmed. And using the developed image processing software, some SPOT P-mode and XS-mode images are merged into the SPOT P＋XS, the high-resolution color composite image.

• The Journal of the Acoustical Society of Korea
• /
• v.40 no.6
• /
• pp.546-554
• /
• 2021

In the signal processing of synthetic aperture sonar, it is subject that the platform in which the sensor array is installed moves along the straight line path. In practical operation in underwater, however, the sensor platform will have trajectory disturbances, diverting from the line path. It causes phase errors in measured signals and then produces deteriorated SAS images. In this study, in order to develop towed SAS, as tools to remove the phase errors associated with the trajectory disturbances of the towfish, motion compensation technique using Redundant Phase Center (RPC) and also Phase Gradient Autofocus (PGA) method is investigated. The performances of these two approaches are examined by means of a simulation for SAS system having a sway disturbance.

• Journal of the Korea Institute of Information and Communication Engineering
• /
• v.16 no.7
• /
• pp.1477-1486
• /
• 2012

In this paper, we propose Multi-view CGH Making System using method of generation of virtual view-point depth image. We acquire reliable depth image using TOF depth camera. We extract parameters of reference-view cameras. Once the position of camera of virtual view-point is defined, select optimal reference-view cameras considering position of it and distance between it and virtual view-point camera. Setting a reference-view camera whose position is reverse of primary reference-view camera as sub reference-view, we generate depth image of virtual view-point. And we compensate occlusion boundaries of virtual view-point depth image using depth image of sub reference-view. In this step, remaining hole boundaries are compensated with minimum values of neighborhood. And then, we generate final depth image of virtual view-point. Finally, using result of depth image from these steps, we generate CGH. The experimental results show that the proposed algorithm performs much better than conventional algorithms.

• Journal of the Korea Institute of Information and Communication Engineering
• /
• v.24 no.12
• /
• pp.1604-1611
• /
• 2020

Light field image captured by a microlens array-based camera has many limitations in practical use due to its low spatial resolution and angular resolution. High spatial resolution images can be easily acquired with a single image super-resolution technique that has been studied a lot recently. But there is a problem in that high angular resolution images are distorted in the process of using disparity information inherent among images, and thus it is difficult to obtain a high-quality angular resolution image. In this paper, we propose light field angular super-resolution that extracts an initial feature map using an dilated convolutional neural network in order to effectively extract the view difference information inherent among images and generates target image using a residual neural network. The proposed network showed superior performance in PSNR and subjective image quality compared to existing angular super-resolution networks.

• Journal of Broadcast Engineering
• /
• v.9 no.3
• /
• pp.257-273
• /
• 2004

In stereoscopic or multi-view three dimensional display systems, the synthesis of intermediate sequences is inevitably needed to assure look-around capability and continuous motion parallax so that it could enhance comfortable 3D perception. The quadtree-based disparity estimation is one of the most remarkable methods for synthesis of Intermediate sequences due to the simplicity of its algorithm and hardware implementation. In this paper, we propose two ideas in order to reduce the annoying flicker at the object boundaries of synthesized intermediate sequences by quadtree-based disparity estimation. First, new split-scheme provides more consistent auadtree-splitting during the disparity estimation. Secondly, adaptive temporal smoothing using correlation between present frame and previous one relieves error of disparity estimation. Two proposed Ideas are tested by using several stereoscopic sequences, and the annoying flickering is remarkably reduced by them.

• The Journal of Korean Institute of Communications and Information Sciences
• /
• v.27 no.6A
• /
• pp.610-617
• /
• 2002

In this paper, a new intermediate view reconstruction technique by using a disparity estimation method based-on the adaptive matching window size is proposed. In the proposed method, once the feature values are extracted from the input stereo image, then the matching window size for the intermediate view reconstruction is adaptively selected in accordance with the magnitude of this feature values. That is, coarse matching is performed in the region having smaller feature values while accurate matching is carried out in the region having larger feature values by comparing with the predetermined threshold value. Accordingly, this new approach is not only able to reduce the mismatching probability of the disparity vector mostly happened in the accurate disparity estimation with a small matching window size, but is also able to reduce the blocking effect occurred in the disparity estimation with a large matching window size. Some experimental results on the 'Parts' and 'Piano' images show that the proposed method improves the PSNR about 2.32∼4.16dB and reduces the execution time to about 39.34∼65.58% than those of the conventional matching methods.

• Journal of the Korean Society of Radiology
• /
• v.15 no.5
• /
• pp.613-620
• /
• 2021

The purpose of this study is to propose a convolutional neural network model that can classify normal and abnormal(cardiomegaly) in chest X-ray images. The training data and test data used in this paper were used by acquiring chest X-ray images of patients diagnosed with normal and abnormal(cardiomegaly). Using the proposed deep learning model, we classified normal and abnormal(cardiomegaly) images and verified the classification performance. When using the proposed model, the classification accuracy of normal and abnormal(cardiomegaly) was 99.88%. Validation of classification performance using normal images as test data showed 95%, 100%, 90%, and 96% in accuracy, precision, recall, and F1 score. Validation of classification performance using abnormal(cardiomegaly) images as test data showed 95%, 92%, 100%, and 96% in accuracy, precision, recall, and F1 score. Our classification results show that the proposed convolutional neural network model shows very good performance in feature extraction and classification of chest X-ray images. The convolutional neural network model proposed in this paper is expected to show useful results for disease classification of chest X-ray images, and further study of CNN models are needed focusing on the features of medical images.

• Proceedings of the Korean Society of Broadcast Engineers Conference
• /
• 2013.11a
• /
• pp.169-171
• /
• 2013

최근 Full-HD TV, UHDTV의 보급에 따라 고화질 영상에 대한 수요가 증가하고 있으며 N-Screen 서비스의 확장으로 고화질 영상을 빠르게 전송하는 문제의 중요성은 더욱 커지고 있다. 고화질 영상을 빠르게 전송하기 위해서는 압축 효율의 향상이 필요한데, 일반적으로 영상에 잡음이 많을 때에는 압축 효율이 떨어진다. 본 논문에서는 다양한 원인의 잡음들 중에 film grain noise에 초점을 맞추어 이를 조절하여 영상압축의 효율을 높이는 방법을 연구한다. film grain은 영화촬영 방법 및 환경 등에 따라 강도가 달라지기도 하지만 필름으로 촬영한 모든 영화에서 쉽게 관찰할 수 있으며 앞으로도 계속 포함이 될 것으로 예상되고, 디지털 영화의 경우에도 저조도에서는 이와 비슷한 특성의 잡음이 발생한다. 재안하는 방법에서는 film grain이 포함된 영상에서 grain을 추출/제거한 영상을 압축하며 추출한 film grain에서 작은 영역을 선택하여 sample grain을 만든 후 별도로 압축한다. 디코더에서 grain을 없앤 영상만을 보여줄 수 있지만, 경우에 따라 grain이 없으면 심미적으로 오히려 좋지 않은 결과가 보이기도 한다. 따라서 압축을 푼 후에는 sample grain에서 원본 영상 크기의 grain을 합성한 후 grain을 제거한 영상과 더하여 grain이 포함된 영상을 재 생성한다. 실험한 결과 원본과 유사한 grain이 생성되면서 압축효율이 향상됨을 확인할 수 있다.

• Proceedings of the KIEE Conference
• /
• 2000.07d
• /
• pp.3236-3237
• /
• 2000

본 논문에서는 의료 영상분야에서 많이 활용되고 있는 기존의 영상 획득,저장 및 전송 시스템(PACS)에 스테레오 내시경 영상이 추가될 수 있도록 기존의 PACS에 스테레오 기능의 관찰(viewing) 시스템을 추가하였으며 기본 기능으로 1) 3차원 스테레오 좌,우 독립 영상의 선택과 합성 2) 기존의 합성된 스테레오 영상의 선택이 가능하도록 하였으며 3) 스테레오 영상의 Dicom 표준이 없는 상황을 고려하여 기본적인 카메라 관련 사항(카메라 사양, 초점 거리, 베이스라인 등)을 입력할 수 있는 기록 필드를 삽입하였다. 또한 임상적으로 수용 가능한 3차원 스테레오 영상의 PACS내 효율적 저장법을 제시하기 위하여 의료 영상에 많이 활용되는 JPEG과 Wavelet 압축법을 각각 이용하여 3차원 좌,우 독립영상과 복합 영상의 효율적 압축비를 PSNR을 중심으로 비교하였다.