• The Journal of the Korea institute of electronic communication sciences
• /
• v.5 no.2
• /
• pp.152-157
• /
• 2010

The development of computer and internet has introduced various types of media - such as, image, audio, video, and voice - to the traditional text-based information. However, most of the information retrieval systems are based only on text, which results in the absence of ability to use available information. By utilizing the available media, one can improve the performance of search system, which is commonly called content-based retrieval and content-based image retrieval system specifically tries to incorporate the analysis of images into search systems. In this paper, a content-based image retrieval system using HSI color space, ART2 algorithm, and SOM algorithm is introduced. First, images are analyzed in the HSI color space to generate several sets of features describing the images and an SOM algorithm is used to provide candidates of training features to a user. The features that are selected by a user are fed to the training part of a search system, which uses an ART2 algorithm. The proposed system can handle the case in which an image belongs to several groups and showed better performance than other systems.

• Journal of the Microelectronics and Packaging Society
• /
• v.30 no.4
• /
• pp.69-78
• /
• 2023

With the rapid growth of artificial intelligence, the demand for semiconductors is enormously increasing everywhere. To ensure the manufacturing quality and quantity simultaneously, the importance of automatic defect detection during the packaging process has been re-visited by adapting various deep learning-based methodologies into automatic packaging defect inspection. Deep learning (DL) models require a large amount of data for training, but due to the nature of the semiconductor industry where security is important, sharing and labeling of relevant data is challenging, making it difficult for model training. In this study, we propose a new framework for securing sufficient data for DL models with fewer computing resources through a divide-and-conquer approach. The proposed method divides high-resolution images into pre-defined sub-regions and assigns conditional labels to each region, then trains individual sub-regions and boundaries with boundary loss inducing the globally coherent and seamless images. Afterwards, full-size image is reconstructed by combining divided sub-regions. The experimental results show that the images obtained through this research have high efficiency, consistency, quality, and generality.

• Journal of the Korea Institute of Information and Communication Engineering
• /
• v.23 no.9
• /
• pp.1049-1059
• /
• 2019

We propose a learning-based image segmentation algorithm. Starting from super-pixels, our method learns the probability of merging two regions based on the ground truth made by humans. The learned information is used in determining whether the two regions should be merged or not in a segmentation stage. Unlike exiting learning-based algorithms, we use both local and object information. The local information represents features computed from super-pixels and the object information represent high level information available only in the learning process. The object information is considered as privileged information, and we can use a framework that utilize the privileged information such as SVM+. In experiments on the Berkeley Segmentation Dataset and Benchmark (BSDS 500) and PASCAL Visual Object Classes Challenge (VOC 2012) data set, out model exhibited the best performance with a relatively small training data set and also showed competitive results with a sufficiently large training data set.

• Progress in Medical Physics
• /
• v.20 no.2
• /
• pp.62-71
• /
• 2009

Recently, optical coherence tomography (OCT) has demonstrated considerable promise for the noninvasive assessment of biological tissues. However, OCT images difficult to analyze due to speckle noise. In this paper, we tested various image processing techniques for speckle removal of human and rabbit cartilage OCT images. Also, we distinguished the images which get with methods of image segmentation for OCT images, and found the most suitable method for segmenting an image. And, we selected image segmentation suitable for OCT before image reconstruction. OCT was a weak point to system design and image processing. It was a limit owing to measure small a distance and depth size. So, good edge matching algorithms are important for image reconstruction. This paper presents such an algorithm, the chamfer matching algorithm. It is made of background for 3D image reconstruction. The purpose of this paper is to describe good image processing techniques for speckle removal, image segmentation, and the 3D reconstruction of cartilage OCT images.

• Proceedings of the Korean Information Science Society Conference
• /
• 1999.10a
• /
• pp.296-298
• /
• 1999

컬러 히스토그램은 멀티미디어 이미지 데이터의 특성을 표현하기 위하여 널이 이용되어 왔다. 그러나 컬러 히스토그램을 고차원으로 설정할 경우 색인 구조에 효율적이지 못할 뿐만 아니라 유사도 계산에서도 고비용이 요구된다. 이러한 단점을 보완하기 위해 히스트그램의 차원을 줄이는 여러 방법이 제시되어 왔으나 이미지의 색상정보 손실을 피할 수 없으며, 이미지의 전체 히스토그램으로는 이미지의 레이아웃을 고려할 수 없기 때문에 필터링을 통한 후보 선정 시 상이한 이미지가 선택되어지는 문제점을 지닌다. 본 논문에서는 이미지를 일정한 크기의 타일로 분할한 이미지 타일 평균 RGB 방법을 제안하였으며, 실험을 통하여 제안한 방법의 성능을 평가하였다.

• Proceedings of the Korea Multimedia Society Conference
• /
• 2001.06a
• /
• pp.185-188
• /
• 2001

의료 영상 처리 기술은 질병의 진단 및 치료를 위한 계획이나 방법을 결정하는데 있어 매우 중요한 역할을 하고 있다. 뇌 MR 영상에서의 질병 진단을 위한 전처리 단계로서 필수적으로 이루어져야 하는 단계가 영상 분할 단계이다. 본 논문에서는 뇌의 질병 진단에 사용할 수 있는 자료를 제공하기 위한 뇌 영상 분할 방법을 제시한다. T2 강조 영상의 반전된 영상에서 원본 영상을 뺀 차이 영상의 결과로 회백질·뇌척수액·비정상 영역이 두드러지게 나타나는 점을 이용해 회백질 뇌척수액·비정상 영역과 백질 영역을 분리하는 방법을 제안한다. 또한 뇌척수액 영역의 위치 정보와 몇 가지 특징들을 정의하여 분할되어진 회백질·뇌척수액· 비정상 영역에서 뇌척수액 영역만을 분할하는 방법을 제시한다. 600 여 개의 T2 강조 영상에 대해서 실험을 행하러 비교적 정확한 분할 결과를 유도할 수 있었음을 확인하였다.

• Journal of Korea Multimedia Society
• /
• v.15 no.4
• /
• pp.439-448
• /
• 2012

TFT-LCD panel images have non-uniform brightness, noise signal and defect signal. It is hard to divide defect signal because of non-uniform brightness and noise signal, so various divide methods have being developed. In this paper, we suggest method to divide defective regions on TFT-LCD panel image by estimating a menas of two different size of windows, which is suggested by Eikvil et al., and using difference of them. But in this method, the size of detectable defects is restricted by the size of window, hence it has inefficient problem that the size of window have to increase to divide a large defect region. To solve this problem we suggest an algorithm which can divide various size of defects, by using Multi-scale and restrict a detectable size of defects in each scale. To prove an efficiency of suggested algorithm, we show that resulting images of real TFT-LCD panel images and an artificial image with various defects.

• Proceedings of the Korean Institute of Information and Commucation Sciences Conference
• /
• 2021.05a
• /
• pp.81-83
• /
• 2021

In this paper, we propose a method to segment a plant from a plant image using U-Net. The network is an end-to-end fully convolutional network that is mainly used for image segmentation. When training the network, we used a binary image that is acquired by the manual segmentation of a plant from the background. Experimental results show that the U-Net based segmentation network can extract a plant from a digital image accurately.

• Journal of the Institute of Convergence Signal Processing
• /
• v.12 no.3
• /
• pp.169-175
• /
• 2011

In this paper, an automatic scoring system of the OMR answer sheet is implemented using Gray Scale and image segmentation method. The proposed method was used to extract the OMR data on multiple-choice answer sheet from captured image. In addition, On-line scoring system is developed and implemented to mark the short-answer type on the reverse side. Therefore, teachers can mark the short-answer type for anytime and anywhere within the available time. There were many advantages to mark of the multiple-choice answer sheet without additional OMR reader. In the future, the grading of short-answer type will be more efficient if it were performed by using an automatic scoring system based on image processing.

• Proceedings of KOSOMES biannual meeting
• /
• 2005.11a
• /
• pp.203-208
• /
• 2005

The CGO (Chip on Glass) to be measured a few micro unit is captured by line scan camera for the accuracy of chip inspection. But it is very sensitive to scan speed and lighting conditions. In this paper, we propose the methods to increase the accuracy of faulty detection by image subtraction. Image subtraction is detected faultiness by subtracting the image of a ' perfect ' COG from trot of the sample under tests. For image subtraction to be successful, the two images must be pre챠sely registered The two images is registered by the area segmentation pattern matching, and the result image get by operating the gradient mask image and the image to practice subtraction. A series of experimentation showed that the proposed algorithm shows substantial improvement over the other image subtraction methods.