• KSII Transactions on Internet and Information Systems (TIIS)
• /
• 제15권7호
• /
• pp.2547-2567
• /
• 2021

It has been widely acknowledged that occlusion impairments adversely distress many face recognition algorithms' performance. Therefore, it is crucial to solving the problem of face image occlusion in face recognition. To solve the image occlusion problem in face recognition, this paper aims to automatically de-occlude the human face majority or discriminative regions to improve face recognition performance. To achieve this, we decompose the generative process into two key stages and employ a separate generative adversarial network (GAN)-based network in both stages. The first stage generates an initial coarse face image without an occlusion mask. The second stage refines the result from the first stage by forcing it closer to real face images or ground truth. To increase the performance and minimize the artifacts in the generated result, a new refine loss (e.g., reconstruction loss, perceptual loss, and adversarial loss) is used to determine all differences between the generated de-occluded face image and ground truth. Furthermore, we build occluded face images and corresponding occlusion-free face images dataset. We trained our model on this new dataset and later tested it on real-world face images. The experiment results (qualitative and quantitative) and the comparative study confirm the robustness and effectiveness of the proposed work in removing challenging occlusion masks with various structures, sizes, shapes, types, and positions.

• 대한전기학회:학술대회논문집
• /
• 대한전기학회 2005년도 심포지엄 논문집 정보 및 제어부문
• /
• pp.60-62
• /
• 2005

In this study, the automatic tracking algorithm tracing a human face is proposed by using YCbCr color coordinated information and its thermal properties expressed in terms of thermal indexes in an infra-red image. The facial candidates are separately estimated in CbCr color and infra-red domain, respectively with applying the morphological image processing operations and the geometrical shape measures for fitting the elliptical features of a human face. The identification of a true face is accomplished by logical 'AND' operation between the refined image in CbCr color and infra-red domain.

• 스마트미디어저널
• /
• 제6권3호
• /
• pp.49-56
• /
• 2017

Images and Videos that include the human face contain a lot of information. Therefore, accurately extracting human face is a very important issue in the field of computer vision. However, in real life, human faces have various shapes and textures. To adapt to these variations, A model-based approach is one of the best ways in which unknown data can be represented by the model in which it is built. However, the model-based approach has its weaknesses when the motion between two frames is big, it can be either a sudden change of pose or moving with fast speed. In this paper, we propose an enhanced human face-tracking model. This approach included human face detection and motion estimation using Cascaded Convolutional Neural Networks, and continuous human face tracking and modeling correction steps using the Active Appearance Model. A proposed system detects human face in the first input frame and initializes the models. On later frames, Cascaded CNN face detection is used to estimate the target motion such as location or pose before applying the old model and fit new target.

• 한국통신학회논문지
• /
• 제17권7호
• /
• pp.767-778
• /
• 1992

3-D 모델 기반 부호화 시스템에서 특징점 추출과 영상합성에 대하여 연구하였다. 얼굴의 특징점들은 영상처리 기술들과 얼굴에 대한 사전지식을 이용하여 자동적으로 추출된다. 추출된 얼굴의 특징점들을 이용하여 얼굴에 정합된 철선 프레임을 특징점의 움직임에 따라 변형시킨다. 변형된 철선 프레임 위에 초기 정면 영상의 질감을 매핑함으로써 합성영상이 만들어진다. 실험결과, 합성영상은 부자연스러움이 거의 나타나지 않았다.

• 인터넷정보학회논문지
• /
• 제8권6호
• /
• pp.155-167
• /
• 2007

사상의학에서는 사람을 네 종류로 구분하며, 한의사들은 종종 이 네 종류에 기반을 두어 특별한 건강 정보와 치료 방법을 제안한다. 얼굴의 특징 비율(표 1)은 사상체질을 판단하는데 있어서 매우 중요한 기준으로 사용되는데, 본 논문에서는 측면얼굴에서 특징 비율을 추출하기 위한 시스템을 제안하였다. 특징 비율을 얻기 위해서는 두 가지를 고려하여야 한다. 하나는 대표 특징들을 선택하는 것이고, 다른 하나는 측면 얼굴 이미지에서 효과적으로 관심 영역을 검출하고, 정확하게 특징 비율을 계산하는 것이다. 논 논문에서 제시한 시스템에서는 적응형 색상 모델을 사용하여 배경에서 측면 얼굴을 분리하였고, 관심 영역 검출을 위해서 기하 모델에 기반한 방법이 사용되었다. 또한 이미지 크기와 머리 포즈에 따른 이미지 변화에 의해서 야기되는 에러 분석을 제시하였다. 제시한 시스템의 성능을 평가하기 위하여 173명의 한국인 왼쪽 얼굴 사진을 이용하여 시스템을 테스트하였고, 정면 사진과 측면 사진을 함께 사용하였을 경우 정면 사진만을 사용한 경우보다 17.99%의 성능 향상을 나타내었다.

• 대한전기학회:학술대회논문집
• /
• 대한전기학회 2006년도 심포지엄 논문집 정보 및 제어부문
• /
• pp.120-121
• /
• 2006

In this paper, We have designed a face recognition system based on the embedded Linux. This paper has an aim in embedded system to recognize the face more exactly. At first, the contrast of the face image is adjusted with lightening compensation method, the skin and lip color is founded based on YCbCr values from the compensated image. To take advantage of the method based on feature and appearance, these methods are applied to the eyes which has the most highly recognition rate of all the part of the human face. For eyes detecting, which is the most important component of the face recognition, we calculate the horizontal gradient of the face image and the maximum value. This part of the face is resized for fitting the eye image. The image, which is resized for fit to the eye image stored to be compared, is extracted to be the feature vectors using the continuous wavelet transform and these vectors are decided to be whether the same person or not with PNN, to miminize the error rate, the accuracy is analyzed due to the rotation or movement of the face. Also last part of this paper we represent many cases to prove the algorithm contains the feature vector extraction and accuracy of the comparison method.

• 한국정보과학회:학술대회논문집
• /
• 한국정보과학회 1998년도 가을 학술발표논문집 Vol.25 No.2 (3)
• /
• pp.765-767
• /
• 1998

Many application in human computer interaction(HCI) require tacking a human face and facial features. In this paper we propose efficient parallel processing system for face tracking under heterogeneous networked. To track a face in the video image we use the skin color information and connected components. In terms of parallelism we choose the master-slave model which has thread for each processes, master and slaves, The threads are responsible for real computation in each process. By placing queues between the threads we give flexibility of data flowing

• 한국통신학회논문지
• /
• 제25권6B호
• /
• pp.1012-1020
• /
• 2000

In this paper, This paper presents a new algorithm which is used for detecting and extracting human masks from a color still image. The regions where each pixel has a value of skin-color were extracted from the Cb and Cr images, after the tone of the color image is converted to YCbCr from. A morphological filter is used to eliminate noise in the resulting image. By scanning it in horizontal and vertical ways under ways under threshold value, first candidate section is chosen. If it is not a face, secondary candidate section is taken and is divided into two candidate sections. The proposed algorithm is not affected by the variation of illuminations, because it uses only Cb and Cr components in YCbCr color format. Moreover, the face recognition was possible regardless of the degree of shifting face, changed shape, various sizes of the face, and the quality of image.

• 한국지능시스템학회논문지
• /
• 제12권2호
• /
• pp.97-102
• /
• 2002

본 논문에서는 CCD 칼라 영상을 이용하여 인간의 감성을 인식할 수 있는 방법을 제안한다. 먼저 CCD 카메라에 의해 획득한 칼라 영상으로부터 피부색 추출 방법을 이용하여 얼굴을 추출한다. 그 다음, 추출된 얼굴 영상으로부터 인간 얼굴의 특징점(눈썹, 눈, 코, 입) 들을 추출하는 방법과 각 특징점들 간의 구조적인 관계로부터 인간의 감성(놀람, 화남, 행복함, 슬픔)을 인식하는 방법을 제안한다. 본 논문에서 제안한 방법은 신경회로망을 이용하여 학습시킴으로써 인간의 감성을 인식한다. 마지막으로, 제안된 방법은 실험을 통해 그 응용 가능성을 확인한다.

• 전기학회논문지P
• /
• 제65권2호
• /
• pp.130-135
• /
• 2016

Gait recognition has advantage of non-contact type recognition. But It has disadvantage of low recognition rate when the pedestrian silhouette is changed due to bag or coat. In this paper, we proposed new method using combination of gait energy image feature and thermal face image feature. First, we extracted a face image which has optimal focusing value using human body rate and Tenengrad algorithm. Second step, we extracted features from gait energy image and thermal face image using linear discriminant analysis. Third, calculate euclidean distance between train data and test data, and optimize weights using genetic algorithm. Finally, we compute classification using nearest neighbor classification algorithm. So the proposed method shows a better result than the conventional method.