• Proceedings of the Korean Information Science Society Conference
• /
• 2002.10d
• /
• pp.508-510
• /
• 2002

본 연구에서는 실시간으로 얼굴을 인식, 검증하기 위한 전처리 단계로써 얼굴을 고속으로 추출하고 추출된 영상을 개선하기 위한 효율적인 기법들을 소개한다. 먼저 RGB로 획득되는 영상을 인간의 시각 구조와 유사한 HSI 컬러 모델로 변환하고 여기서 인간의 피부 영역에 해당하는 컬러 분포를 조사하여 대강의 얼굴 영역을 찾고 이 영역을 대상으로 두 개의 가변 템플릿과의 상관도(Correlation)를 이용하여 최적의 얼굴 안면을 찾는다. 보다 나은 얼굴 인식을 위하여 검출된 얼굴 안면 이미지에서 조명 평면(Illumination plane) 이미지를 추출하여 먼저 불균일성을 보정한 다음 평활화(Equalization)를 수행함으로써 영상을 개선한다.

• Journal of the Korea Institute of Information and Communication Engineering
• /
• v.26 no.9
• /
• pp.1305-1311
• /
• 2022

With the recent rapid development of mobile terminals and personal computers and the advent of neural network technology, real-time face swapping using images has become possible. In particular, the cycle generative adversarial network made it possible to replace faces using uncorrelated image data. In this paper, we propose an input data processing scheme that can improve the quality of face swapping with less training data and time. The proposed scheme can improve the image quality while preserving facial structure and expression information by combining facial landmarks extracted through a pre-trained neural network with major information that affects the structure and expression of the face. Using the blind/referenceless image spatial quality evaluator (BRISQUE) score, which is one of the AI-based non-reference quality metrics, we quantitatively analyze the performance of the proposed scheme and compare it to the conventional schemes. According to the numerical results, the proposed scheme obtained BRISQUE scores improved by about 4.6% to 14.6%, compared to the conventional schemes.

• Proceedings of the Korean Institute of Information and Commucation Sciences Conference
• /
• 2009.10a
• /
• pp.289-292
• /
• 2009

A person does communication between each other using language. But In the case of disabled person can not communication own idea to use writing and gesture. Therefore, In this paper, we embodied communication system using the facial muscle signals so that disabled person can do communication. Especially, After feature extraction of the EEG included facial muscle, it is converted the facial muscle into control signal, and then select character and communicate using a minimum list keyboard.

• Journal of the Korea Institute of Information and Communication Engineering
• /
• v.14 no.6
• /
• pp.1338-1344
• /
• 2010

A person does communication between each other using language. But In the case of disabled person can not communication own idea to use writing and gesture. Therefore, In this paper, we embodied communication system using the facial muscle signals so that disabled person can do communication. Especially, After feature extraction of the EEG included facial muscle, it is converted the facial muscle into control signal, and then select character and communication using a minimum list keyboard.

• Journal of Internet Computing and Services
• /
• v.23 no.4
• /
• pp.35-43
• /
• 2022

Makeup is the most common way to improve a person's appearance. However, since makeup styles are very diverse, there are many time and cost problems for an individual to apply makeup directly to himself/herself.. Accordingly, the need for makeup automation is increasing. Makeup transfer is being studied for makeup automation. Makeup transfer is a field of applying makeup style to a face image without makeup. Makeup transfer can be divided into a traditional image processing-based method and a deep learning-based method. In particular, in deep learning-based methods, many studies based on Generative Adversarial Networks have been performed. However, both methods have disadvantages in that the resulting image is unnatural, the result of makeup conversion is not clear, and it is smeared or heavily influenced by the makeup style face image. In order to express the clear boundary of makeup and to alleviate the influence of makeup style facial images, this study divides the makeup area and calculates the loss function using HoG (Histogram of Gradient). HoG is a method of extracting image features through the size and directionality of edges present in the image. Through this, we propose a makeup transfer network that performs robust learning on edges.By comparing the image generated through the proposed model with the image generated through BeautyGAN used as the base model, it was confirmed that the performance of the model proposed in this study was superior, and the method of using facial information that can be additionally presented as a future study.

• Journal of Korean Society for Geospatial Information Science
• /
• v.24 no.1
• /
• pp.109-119
• /
• 2016

The height datum of Korea is currently separated into land and sea, which makes it difficult to acquire homogeneous and accurate height information throughout the whole nation. In this study, we therefore tried to suggest the more effective way to transform the height information were constructed separately according to each height datum on land and sea to those on the unique height datum using precise geoid models and tidal observations in Korea. For this, Anmyeon island was selected as a study area to develop the precise geoid models based on the height datums land (IMSL) and sea (LMSL), respectively. In order to develop two hybrid geoid models based on each height datum of land an sea, we firstly develop a precise gravimetric geoid model using the remove and restore (R-R) technique with all available gravity observations. The gravimetric geoid model were then fitted to the geometric geoidal heights, each of which is represented as height datum of land or sea respectively, obtained from GPS/Leveling results on 15 TBMs in the study area. Finally, we determined the differences between the two hybrid geoid models to apply the height transformation between IMSL and LMSL. The co-tidal chart model of TideBed system developed by Korea Hydrographic and Oceanographic Agency (KHOA) which was re-gridded to have the same grid size and coverage as the geoid model, in order that this can be used for the height datum transformation from LMSL to local AHHW and/or from LMSL to local DL. The accuracy of height datum transformation based on the strategy suggested in this study was approximately ${\pm}3cm$. It is expected that the results of this study can help minimize not only the confusions on the use of geo-spatial information due to the disagreement caused by different height datum, land and sea, in Korea, but also the economic and time losses in the execution of coastal development and disaster prevention projects in the future.

• Journal of the Korea Society of Computer and Information
• /
• v.15 no.7
• /
• pp.75-81
• /
• 2010

In this paper, we propose an emotion messenger which recognizes face or hand gestures of a user using a WebCAM, converts recognized emotions (joy, anger, grief, happiness) to flash-cones, and transmits them to the counterpart. This messenger consists of face recognition module, hand gesture recognition module, and messenger module. In the face recognition module, it converts each region of the eye and the mouth to a binary image and recognizes wink, kiss, and yawn according to shape change of the eye and the mouth. In hand gesture recognition module, it recognizes gawi-bawi-bo according to the number of fingers it has recognized. In messenger module, it converts wink, kiss, and yawn recognized by the face recognition module and gawi-bawi-bo recognized by the hand gesture recognition module to flash-cones and transmits them to the counterpart. Through simulation, we confirmed that CPU share ratio of the emotion messenger is minimized. Moreover, with respect to recognition ratio, we show that the hand gesture recognition module performs better than the face recognition module.

• Journal of Korean Academy of Oral and Maxillofacial Radiology
• /
• v.24 no.1
• /
• pp.23-28
• /
• 1994

전남대학교 치과대학 학생을 대상으로 평행촬영법으로 촬영한 168장의 구내방사선사진을 디지털 영상으로 변환하여, 필름상의 치석 유무 판독 결과와 모니터상의 판독 결과를 비교하였다. 3명의 구강악안면방사선학 전공 치과의사가 판독을 하되 필름상의 판독 결과를 판정의 기준으로 하였다. 디지털 영상은 상업용 필름-비디오 변환장치, 486DX PC, 그리고 PcVision과 frame grabber을 이용하여 얻었다. 모니터상의 개개 구내방사선사진상은 700×480 pixel로 나타내었다. 디지털 영상의 치석진단의 sensitivity는 0.79-0.90(평균 0.86) 이었고, specificity는 0.03-0.43(평균 0.17) 이었다.

• Proceedings of the Korea Information Processing Society Conference
• /
• 2002.11a
• /
• pp.641-644
• /
• 2002

얼굴 특징점 추출은 현재 많은 연구가 활발히 진행되고 있는 분야로 보안, 인식 등 다양한 응용분야를 갖는다. 본 논문에서는 PC 카메라 및 주민등록증에 있는 사진을 스캔하여 얼굴 특징점을 정확하고 빠른 계산 시간안에 찾을 수 있는 새로운 방법을 제시한다. RGB 색공간을 YUV로 변환하여 Y성분을 히스토그램 균등화 시켜 휘도에 관계없이 얼굴 피부색을 추출한 후 YUV의 V성분을 변형한 V'성분을 이용하여 얼굴의 특징점을 찾는 방법이다. 실험결과 주민등록증 사진과 PC 카메라에서 입력 받은 얼굴 영상이 오류 없이 추출됨이 관찰되었다.

• Proceedings of the Korean Society of Broadcast Engineers Conference
• /
• 2018.06a
• /
• pp.133-135
• /
• 2018

본 논문에서는 포즈에 적응적인 가중치 맵 (weight map)에 기반한, 얼굴 교체시스템을 위한 블렌딩 기법을 제안한다. 우선 얼굴교체를 진행하기 위해 목표얼굴이 들어있는 영상으로부터 실시간으로 얼굴의 기하학적 특징점 (land mark)을 검출한다. 다음 검출된 특징점의 분포에 따라 얼굴영역에 대해 삼각화 (triangulation)를 진행한다. 참조영상에 대해서도 같은 과정을 적용하고 대응되는 영역끼리 워핑 (warping) 변환을 시키면 목표 얼굴과 같은 포즈의 참조얼굴을 얻을 수 있다. 그 다음 두 영상의 피부색 톤을 일치시켜주고 안면교체를 진행한다. 하지만 교체된 영역과 목표 얼굴 사이에 부자연스러운 경계가 발생하게 되는데 블렌딩 기법을 통해 이런 경계를 제거한다. 본 논문에서는 사전에 표준얼굴형태모델을 이용하여 정면 얼굴의 가중치 맵을 생성하고, 표준얼굴형태모델과 목표 얼굴사이 변환관계를 이용하여 포즈에 대응되는 가중치지도를 생성하였다. 이렇게 얻어진 가중치 맵은 일관되게 정해진 가중치 맵에 비해 포즈변화에 적응적으로 대처할 수 있어 보다 자연스러운 얼굴교체 효과를 얻을 수 있다.