• Proceedings of the IEEK Conference
• /
• 2000.06d
• /
• pp.159-162
• /
• 2000

Selecting locally optimum thresholds, based on optimizing a criterion composed of the area variation rate and the compactness of the segmented shape, is presented. The method is shown to have the shape-resolving property in the subtraction image, so that overlapped objects may be resolved into bright and dark evidences characterizing each object. As an application a vehicle detection algorithm robust to the operating conditions could be realized by applying simple merging rules to the geometrically correlated bright and dark evidences obtained by this local thresholding.

• Proceedings of the Korean Information Science Society Conference
• /
• 2003.10b
• /
• pp.553-555
• /
• 2003

사람의 얼굴을 검출 및 인식을 하는 여러 가지 다양한 알고리즘이 소개되고 있다. 본 논문에서는 사람의 피부색을 이용한 컬러정보(Color Information)와 국부 최적 임계치 기법을 사용하여 얼굴의 형상정보를 검출하고 얼굴 영역을 검출하는 방법을 사용한다. 컬러정보를 사용하여 얼굴의 후보영역을 선정한 후에 그 후보영역에서 얼굴의 특징인 눈, 눈썹, 입을 찾는 방법을 제안한다. 피부색은 일정한 분포를 가지고 있기 때문에 후보영역을 비교적 정확히 찾을 수 있으며, 국부 최적 임계치 기법은 효과적인 얼굴 특징 검출방법이다.

• Journal of KIISE:Software and Applications
• /
• v.30 no.7_8
• /
• pp.773-779
• /
• 2003

Locating exact position of facial components is a key preprocessing for realizing highly accurate and reliable face recognition schemes. In this paper, we propose a simple but powerful method for detecting isolated facial components such as eyebrows, eyes, and a mouth, which are horizontally oriented and have relatively dark gray levels. The method is based on the shape-resolving locally optimum thresholding that may guarantee isolated detection of each component. We show that pure facial regions can be determined by grouping facial features satisfying simple geometric constraints on unique facial structure. In the test for over 1000 images in the AR -face database, pure facial regions were detected correctly for each face image without wearing glasses. Very few errors occurred in the face images wearing glasses with a thick frame because of the occluded eyebrow -pairs. The proposed scheme may be best suited for the later stage of classification using either the mappings or a template matching, because of its capability of handling rotational and translational variations.