• 한국통신학회논문지
• /
• 제8권3호
• /
• pp.105-112
• /
• 1983

本論文은 線形空間的不變인 段損(blur)과 自色가우스性雜音에 의해 損傷된 映像에 대한 循環復元(recursive restoration)技法을 論하였다. 映像은 確率設計學的으로 그 平均과 相關函數(correlation function)에 의해 特徵지워진다. 隣接모델(neighborhood model)에 指數的自己相關函數(exponential autocorrelation function)가 사용되며 解析이 간단하고 편리하므로 映像度相關函數를 나타내는데 벡터 모델이 사용된다. 이 벡터 모델을 基本으로 한 映像表現에 있어서 離散的, 統計學的인 12點隣接모델이 開發되고 次元의 增加를 抑制하며 破損되고 雜音섞인 映像을 復元하기 위한 窓(window)移動處理技法이 使用되었다. 12點隣接모델 8點隣接모델보다 優秀한 것으로 나타나며 隣接의 많은 畵素를 요하는 精密畵像에 適合함을 보인다. 이 結果는 線形필터링을 요하는 映像處理에 널리 이용될 수 있음을 나타낸다.

• KSII Transactions on Internet and Information Systems (TIIS)
• /
• 제17권5호
• /
• pp.1433-1449
• /
• 2023

In recent years, the rapid development of social networks has led to a rapid increase in the amount of information available on the Internet, which contains a large amount of sensitive information related to pornography, politics, and terrorism. In the aspect of sensitive image detection, the existing machine learning algorithms are confronted with problems such as large model size, long training time, and slow detection speed when auditing and supervising. In order to detect sensitive images more accurately and quickly, this paper proposes a multiclassification sensitive image detection method based on lightweight Convolutional Neural Network. On the basis of the EfficientNet model, this method combines the Ghost Module idea of the GhostNet model and adds the SE channel attention mechanism in the Ghost Module for feature extraction training. The experimental results on the sensitive image data set constructed in this paper show that the accuracy of the proposed method in sensitive information detection is 94.46% higher than that of the similar methods. Then, the model is pruned through an ablation experiment, and the activation function is replaced by Hard-Swish, which reduces the parameters of the original model by 54.67%. Under the condition of ensuring accuracy, the detection time of a single image is reduced from 8.88ms to 6.37ms. The results of the experiment demonstrate that the method put forward has successfully enhanced the precision of identifying multi-class sensitive images, significantly decreased the number of parameters in the model, and achieved higher accuracy than comparable algorithms while using a more lightweight model design.

• 정보처리학회논문지:소프트웨어 및 데이터공학
• /
• 제12권12호
• /
• pp.519-524
• /
• 2023

해싱 기반 이미지 검색에서는 조작된 이미지의 해시코드가 원본 이미지와 달라 동일한 이미지 검색이 어렵다. 본 논문은 이미지의 질감, 모양, 색상 등 특징 정보로부터 지각적 해시코드를 생성하는 자기 감독 기반 딥해싱 모델을 제안하고 평가한다. 비교 모델은 오토인코더 기반 변분 추론 모델들이며, 인코더는 완전 연결 계층, 합성곱 신경망과 트랜스포머 모듈 등으로 설계된다. 제안된 모델은 기하학적 패턴을 추출하고 이미지 내 위치 관계를 활용하는 SimAM 모듈을 포함하는 변형 추론 모델이다. SimAM은 뉴런과 주변 뉴런의 활성화 값을 이용한 에너지 함수를 통해 객체 또는 로컬 영역이 강조된 잠재 벡터를 학습할 수 있다. 제안 방법은 표현 학습 모델로 고차원 입력 이미지의 저차원 잠재 벡터를 생성할 수 있으며, 잠재 벡터는 구분 가능한 해시코드로 이진화 된다. CIFAR-10, ImageNet, NUS-WIDE 등 공개 데이터셋의 실험 결과로부터 제안 모델은 비교 모델보다 우수하며, 지도학습 기반 딥해싱 모델과 동등한 성능이 분석되었다.

• 대한원격탐사학회:학술대회논문집
• /
• 대한원격탐사학회 2005년도 Proceedings of ISRS 2005
• /
• pp.91-94
• /
• 2005

A numerical formula that presents relationship between a point of a satellite image and its ground position is called a sensor model. For precise geolocation of satellite images, we need an error-free sensor model. However, the sensor model based on GOES ephemeris data has some error, in particular after Image Motion Compensation (IMC) mechanism has been turned off. To solve this problem, we investigate three sensor models: Collinearity model, Direct Linear Transform (DLT) model and Orbit-based model. We apply matching between GOES images and global coastline database and use successful results as control points. With control points we improve the initial image geolocation accuracy using the three models. We compare results from three sensor models that are applied to GOES-9 images. As a result, a suitable sensor model for precise geolocation of GOES-9 images is proposed.

• 한국실내디자인학회논문집
• /
• 제16호
• /
• pp.161-167
• /
• 1998

The Study to verify the relations between the store choice behavior and the interior environmental image. the point of view is that the image of store affects the behavior of customers when they choose the store for shopping. The method used for this study is library survey for constructing the store choice model and the field survey for veryfying the model. The store choice behavior of customers occurs according to the procedure : environment exposure-development of cognitive structure-image developing by the emotional response-attitude decision- purchasing. The image of a store is one of the very important elements in marketing stragy. The group which showed the affirmative response to the emvironmental design aspects also showed the high affirmative response to the store image. The group which showed consistent response to the image of store and the image of him/herself showed more affirmative response to the environmental design aspects. Good image to a store affects affirmative attitude of customers and it consequently the store choice behavior

• 한국멀티미디어학회논문지
• /
• 제21권2호
• /
• pp.158-165
• /
• 2018

Since a large amount of multimedia image has dramatically increased, it is important to search semantically relevant image. Thus, several semantic image annotation methods using RDF(Resource Description Framework) model in mobile environment are introduced. Earlier studies on annotating image semantically focused on both the image tag and the context-aware information such as temporal and spatial data. However, in order to fully express their semantics of image, we need more annotations which are described in RDF model. In this paper, we propose an annotation method inferencing with RDFS entailment rules and user defined rules. Our approach implemented in Moment system shows that it can more fully represent the semantics of image with more annotation triples.

• 제어로봇시스템학회:학술대회논문집
• /
• 제어로봇시스템학회 2005년도 ICCAS
• /
• pp.1244-1248
• /
• 2005

In this paper, we propose fuzzy c-means (FCM) to solve recognition errors in invariant range image, multi-pose face recognition. Scale, center and pose error problems were solved using geometric transformation. Range image face data was digitized into range image data by using the laser range finder that does not depend on the ambient light source. Then, the digitized range image face data is used as a model to generate multi-pose data. Each pose data size was reduced by linear reduction into the database. The reduced range image face data was transformed to the gradient face model for facial feature image extraction and also for matching using the fuzzy membership adjusted by fuzzy c-means. The proposed method was tested using facial range images from 40 people with normal facial expressions. The output of the detection and recognition system has to be accurate to about 93 percent. Simultaneously, the system must be robust enough to overcome typical image-acquisition problems such as noise, vertical rotated face and range resolution.

• 대한전자공학회논문지
• /
• 제17권2호
• /
• pp.15-22
• /
• 1980

Graphic image data의 효율적인 coding을 위해서, 한글 image와 영문 image의 통계적인 특성들이 측정 비교되었다. 또한 Markov model에 의한 run length의 확률분포와 측정된 run length의 확률분포가 비교 검토되었다. 측정된 run length 분포는 negative-power 분포에 유사하며, 이것은 한글 image 에서 더욱 뚜렷한 것으로 나타났다. 대표적 네가지 run length code의 coding 특성이 비교되었는데, 영문 image보다 한글image의 coding 특성이 더욱 우수한 것으로 밝혀졌다.

• 한국정보통신학회:학술대회논문집
• /
• 한국정보통신학회 2012년도 추계학술대회
• /
• pp.887-890
• /
• 2012

Image recognition is one of the most important research directions of pattern recognition. Image based road automatic identification technology is widely used in current society, the intelligence has become the trend of the times. This paper studied the image segmentation algorithm theory and its application in road signs recognition system. With the help of image processing technique, respectively, on road signs automatic recognition algorithm of three main parts, namely, image segmentation, character segmentation, image and character recognition, made a systematic study and algorithm. The experimental results show that: the image segmentation algorithm to establish road signs recognition model, can make effective use of smart phone system and application.

• 대한방사선기술학회지:방사선기술과학
• /
• 제45권1호
• /
• pp.41-47
• /
• 2022

The purpose of this study is to apply a deep learning model that can distinguish lung perfusion and lung ventilation images in nuclear medicine, and to evaluate the image classification ability. Image data pre-processing was performed in the following order: image matrix size adjustment, min-max normalization, image center position adjustment, train/validation/test data set classification, and data augmentation. The convolutional neural network(CNN) structures of VGG-16, ResNet-18, Inception-ResNet-v2, and SE-ResNeXt-101 were used. For classification model evaluation, performance evaluation index of classification model, class activation map(CAM), and statistical image evaluation method were applied. As for the performance evaluation index of the classification model, SE-ResNeXt-101 and Inception-ResNet-v2 showed the highest performance with the same results. As a result of CAM, cardiac and right lung regions were highly activated in lung perfusion, and upper lung and neck regions were highly activated in lung ventilation. Statistical image evaluation showed a meaningful difference between SE-ResNeXt-101 and Inception-ResNet-v2. As a result of the study, the applicability of the CNN model for lung scintigraphy classification was confirmed. In the future, it is expected that it will be used as basic data for research on new artificial intelligence models and will help stable image management in clinical practice.