• Journal of the Korea Computer Graphics Society
• /
• v.9 no.1
• /
• pp.11-18
• /
• 2003

본 논문에서는 회화적 렌더링에서 칼라변환을 이용한 브러쉬 스트로크의 생성에 관한 새로운 알고리즘을 제안한다. 본 논문의 브러쉬 스트로크 생성을 위한 전체적인 구성은 다음과 같다. 첫째, 두 장의 사진(한 장의 소스 이미지와 한 장의 참조 이미지)을 입력으로 하여 칼라 변환 이론을 적용하여 색상 테이블이 바뀐 새로운 이미지를 생성한다. 이 방법은 소스 이미지의 칼라 분포 형태를 창조 이미지의 칼라 분포 형태로 변환하기 위해, 선형 히스토그램 매칭이라 불리는, 간단한 통계학적 방법을 이용한다. 둘째, 가우시안 블러링과 소벨 필터를 이용하여 에지를 검출한다. 검출된 에지는 브러쉬 스트로크 렌더링 시 에지 부분에서 스트로크를 클리핑 함으로써 이미지의 윤곽선 보존을 위해 사용된다. 셋째, 브러쉬 스트로크의 방향을 결정하기 위한 방향맵을 생성한다. 방향맵은 입력 영상에 대한 영역 분할 및 병합을 토대로 만들어진다. 영역별 각 픽셀들에 대해 이미지 그래디언트에 기초한 일정한 방향을 부여함으로써 방향맵을 구성한다. 넷째, 구성된 방향맵을 참조하여 브러쉬 스트로크 생성의 기초가 되는 베지어 곡선(Bezier Curve)의 제어점(Control point)을 설정한다. 실제 회화작품에서 사용되는 브러쉬 스트로크는 일반적으로 곡선의 형태를 이루므로 곡선 표현이 가능한 베지어 곡선을 이용하여 브러쉬 스트로크를 표현하였다. 마지막으로, 생성된 브러쉬 스트로크를 에지부문에서 클리핑하고 배경색을 참조하여 블렌딩하거나 퐁 조명 모델을 이용하여 이미지에 적용하게 된다.

• Proceedings of the Korea Multimedia Society Conference
• /
• 2012.05a
• /
• pp.4-5
• /
• 2012

원거리에서의 얼굴 식별은 낮은 영상 해상도를 비롯하여 블러와 잡음으로 인한 어려움이 크다. 더욱이 모바일 장치에서 실시간 처리를 하기 위하여 느린 수행속도와 제한된 메모리 등 모바일 계산환경을 필히 고려하여야 한다. 본 논문은 모바일 환경에서 주파수 영역의 정합 필터를 이용한 얼굴 식별 방법을 제안한다. 얼굴 식별은 선형(linear) 및 위상(phase-only) 필터, 순차적인 검증 단계를 이용하여 수행된다. 얼굴 후보 윈도우 영역은 선형 필터와 위상 필터를 수행하여 검출하고 순차적인 검증 단계는 피부색 테스트와 경계 마스크 필터링 테스트로 구성한다. 제안된 방법은 Android 플랫폼에서 Java을 이용하여 모바일 폰에서 개발하였다. 예비실험 결과는 모바일 환경에서 얼굴 식별이 실시간으로 성공적으로 수행될 수 있음을 보인다.

• Journal of the Institute of Electronics and Information Engineers
• /
• v.54 no.6
• /
• pp.72-79
• /
• 2017

In this paper, we present a new saliency detection algorithm using the mutual information of wavelet subbands. Our method constructs an intermediate saliency map using the power operation and Gaussian blurring for high-frequency wavelet coefficients. After combining three intermediate saliency maps according to the direction of wavelet subband, we find the main directional components using entropy measure. The amount of mutual information of each subband is obtained centering on the subband having the minimum entropy The final saliency map is detected using Minkowski sum based on weights calculated by the mutual information. As a result of the experiment on CAT2000 and ECSSD databases, our method showed good detection results in terms of ROC and AUC with few computation times compared with the conventional methods.

• Journal of the Korea Academia-Industrial cooperation Society
• /
• v.21 no.6
• /
• pp.84-90
• /
• 2020

As the uploading and downloading of data through the Internet is becoming more common, data including personal information are easily exposed to unauthorized users. In this study, we detect a target area in images that contain personal information, except for the background, and we protect the detected target area by using a blocking method suitable for the surrounding situation. In this method, only the target area from color image input containing personal information is segmented based on skin color. Subsequently, blurring of the corresponding area is performed in multiple stages based on the surrounding situation to effectively block the detected area, thereby protecting the personal information from being exposed. Experimental results show that the proposed method blocks the object region containing personal information 2.3% more accurately than an existing method. The proposed method is expected to be utilized in fields related to image processing, such as video security, target surveillance, and object covering.

• Journal of the Korea Academia-Industrial cooperation Society
• /
• v.21 no.10
• /
• pp.454-460
• /
• 2020

In recent years, the rapid increases in video distribution and viewing over the Internet have increased the risk of personal information exposure. In this paper, a method is proposed to robustly identify areas in images where a person's privacy is compromised and simultaneously blocking the object area by blurring it while rapidly tracking it using a prediction algorithm. With this method, the target object area is accurately identified using artificial neural network-based learning. The detected object area is then tracked using a location prediction algorithm and is continuously blocked by blurring it. Experimental results show that the proposed method effectively blocks private areas in images by blurring them, while at the same time tracking the target objects about 2.5% more accurately than another existing method. The proposed blocking method is expected to be useful in many applications, such as protection of personal information, video security, object tracking, etc.

• Journal of the Korea Institute of Information and Communication Engineering
• /
• v.7 no.4
• /
• pp.758-766
• /
• 2003

This paper proposes a tree-based hierarchical image mosaicing system using camera and object parameters for efficient video database construction. Gray level histogram difference and average intensity difference are proposed for scene change detection of input video. Camera parameter measured by utilizing least sum of square difference and affine model, and difference image is used for similarity measure of two input images. Also, dynamic objects are searched by through macro block setting and extracted by using region splitting and 4-split detection methods. Dynamic trajectory evaluation function is used for expression of dynamic objects, and blurring is performed for construction of soft and slow mosaic image.

• Journal of the Korean Society for Railway
• /
• v.13 no.3
• /
• pp.283-289
• /
• 2010

Recent advance of digital camera results in that image signal processing techniques are widely adopted to railroad security management. However, due to the nature of railroad management many images are acquired in low light level environment such as night scenes. The lack of light causes lots of noise in the image, which degrades image quality and causes errors in the next processes. 3D noise reducing techniques produce better results by using consecutive sequence of images. On the other hand, they cause degradation such as motion blur if there are motions in the sequence. In this paper, we use an adaptive weight filter to estimate more accurate motions and use the result of the adaptive filter to 3D result to improve objective and subjective mage quality.

• Journal of the Korea Society of Computer and Information
• /
• v.26 no.2
• /
• pp.19-25
• /
• 2021

In this paper proposes a method of finding missing persons based on face-recognition technology and deep learning. In this paper, a real-time face-recognition technology was developed, which performs face verification and improves the accuracy of face identification through data fortification for face recognition and convolutional neural network(CNN)-based image learning after the pre-processing of images transmitted from a mobile device. In identifying a missing person's image using the system implemented in this paper, the model that learned both original and blur-processed data performed the best. Further, a model using the pre-learned Noisy Student outperformed the one not using the same, but it has had a limitation of producing high levels of deflection and dispersion.

• Proceedings of the Korean Institute of Information and Commucation Sciences Conference
• /
• 2013.05a
• /
• pp.364-367
• /
• 2013

In this paper, we propose a method about reducing facial blemishes using Morphological Operation. First, we detect skin region using pixel data of RGB's each channel image. we create histogram of skin region R, G, B channel and save 3 pixel values that are high frequency pixel value in each channel. After than, we find facial blemishes using Black-hat operation. The pixel value of facial blemishes changes average of its pixel value, 8-neighborhood pixel value and high frequency pixel values. And the facial blemishes pixel is blurred with median filter. The result of this test with facial pictures that have facial blemishes, we prove that this system that correct the face skin using reduction facial Blemishes is more efficient method than correct the face skin just using lighting up.

• Journal of the Korea Society of Computer and Information
• /
• v.27 no.5
• /
• pp.47-54
• /
• 2022

Recently, with the development of deep learning technology, research on recognizing human behavior is in progress. In this paper, a study was conducted to recognize risky behaviors that may occur in a single-person household environment using deep learning technology. Due to the nature of single-person households, personal privacy protection is necessary. In this paper, we recognize human dangerous behavior in privacy protection video with Gaussian blur filters for privacy protection of individuals. The dangerous behavior recognition method uses the YOLOv5 model to detect and preprocess human object from video, and then uses it as an input value for the behavior recognition model to recognize dangerous behavior. The experiments used ResNet3D, I3D, and SlowFast models, and the experimental results show that the SlowFast model achieved the highest accuracy of 95.7% in privacy-protected video. Through this, it is possible to recognize human dangerous behavior in a single-person household environment while protecting individual privacy.