DC Field | Value | Language |
---|---|---|
dc.contributor.advisor | Ro, Yong-Man | - |
dc.contributor.advisor | 노용만 | - |
dc.contributor.author | Kim, Hyung-Il | - |
dc.contributor.author | 김형일 | - |
dc.date.accessioned | 2015-04-23T06:13:57Z | - |
dc.date.available | 2015-04-23T06:13:57Z | - |
dc.date.issued | 2013 | - |
dc.identifier.uri | http://library.kaist.ac.kr/search/detail/view.do?bibCtrlNo=566522&flag=dissertation | - |
dc.identifier.uri | http://hdl.handle.net/10203/196694 | - |
dc.description | 학위논문(석사) - 한국과학기술원 : 전기및전자공학과, 2013.8, [ iii, 38 p. ] | - |
dc.description.abstract | Automatic face recognition (FR) has long been established as one of the most active research areas in computer vision and image understanding. In well-controlled condition, FR technologies have been extensive-ly studied and are relatively matured. However, accurate FR in unconstrained condition is still a tough task due to faces images containing severe variations in appearance. Of variations, the varying resolution problem by a distance taking a picture is serious problem. For example, in surveillance system with resolution con-straints, a critical and practical issue is that the distance from subject to the camera during enrollment is dif-ferent from that utilized for verification or identification. Another example is smart camera that has a function of automatic power-saving contingent upon the importance of event controls the camera resolution. From the above examples, we can consider a FR scenario at which probe face images have lower resolution compared to that of enrolled face image in training data. Such a FR scenario may lead to significant degradation in recognition performance because of lacking discrimination information in probe face images and mismatch problem. However, the previous work has focused on solving low resolution problem. These approaches are mainly computationally intensive method due to the optimization problem even if there has been the require-ment of low computing power in practical FR. In this thesis, we proposed the new face recognition (FR) algo-rithm feasible for use of practical FR applications. Specifically, the proposed FR aims to be effective in the challenges of varying face resolution while satisfying low computing power which is often addressed in low-powered devices. By introducing the “feature scalability”, we flexibly take a feature that corresponds to lower face resolution for the purpose of sufficiently high resolution of training face given a feature of training face image of sufficiently high resolution. Because... | eng |
dc.language | eng | - |
dc.publisher | 한국과학기술원 | - |
dc.subject | Face Recognition | - |
dc.subject | 제약되지 않은 해상도 | - |
dc.subject | 특징 추출 | - |
dc.subject | 특징추출의 확장성 | - |
dc.subject | 얼굴인식 | - |
dc.subject | Unconstrained Resolution | - |
dc.subject | Feature Scalability | - |
dc.subject | Feature Extraction | - |
dc.title | Face recognition in unconstrained spatial resolution using feature scalability | - |
dc.title.alternative | 제약되지 않은 영상 해상도 환경에서 특징추출의 확장성을 이용한 얼굴인식에 관한 연구 | - |
dc.type | Thesis(Master) | - |
dc.identifier.CNRN | 566522/325007 | - |
dc.description.department | 한국과학기술원 : 전기및전자공학과, | - |
dc.identifier.uid | 020114346 | - |
dc.contributor.localauthor | Ro, Yong-Man | - |
dc.contributor.localauthor | 노용만 | - |
Items in DSpace are protected by copyright, with all rights reserved, unless otherwise indicated.