专利快速检索-快速检索全球专利，免费商用专利数据库-IPRDB

1. 发明授权

KR100893086B1 조명 변화에 강인한 얼굴 검출 방법 失效
标题翻译：用于检测面部稳定到照明变化的方法
公开(公告)号：KR100893086B1
公开(公告)日：2009-04-14
申请号：KR1020060027773
申请日：2006-03-28
申请人： (주)코아정보시스템 , 김대진
发明人： 강태운 , 이승진 , 김대진 , 전봉진
IPC分类号： G06T7/60 , G06T7/00
摘要： 본 발명에 따른 조명 변화에 강인한 얼굴 검출 방법은,

where,
{ : 픽셀 주변의 3*3 영역의 픽셀의 평균값,
: ( )위치의 픽셀값}
입력 영상에 대해서 상기 수식에 의해 census transform을 적용하여 특징(feature) 값을 추출하는 제 1단계; 추출된 특징(feature)값을 사용하여 얼굴 검출 학습을 수행함으로 표준 특징 값 그룹인 룩업테이블을 얻어내는 제 2단계; 상기 제 2단계로부터 얻어진 룩업 테이블의 특징 값을 이용하여 실제 얼굴 검출을 수행하는 3단계;로 이루어진 것을 특징으로 한다.
본 발명에 따른 조명 변화에 강인한 얼굴 검출 방법에 따르면, 보다 정밀하게 얻어진 룩업 테이블을 통해 검출 속도를 단축하면서 조명 변화에 따른 얼굴 영상을 탁월하게 검출할 수 있는 장점이 있다.

2. 发明公开

KR1020070097636A 조명 변화에 강인한 얼굴 검출 방법 失效
标题翻译：用于检测面部稳定到照明变化的方法
公开(公告)号：KR1020070097636A
公开(公告)日：2007-10-05
申请号：KR1020060027773
申请日：2006-03-28
申请人： (주)코아정보시스템 , 김대진
发明人： 강태운 , 이승진 , 김대진 , 전봉진
IPC分类号： G06T7/60 , G06T7/00
CPC分类号： G06K9/00221 , G06K9/46 , G06T7/40
摘要： A face detection method which is strong in an illumination change is provided to reduce a calculation quantity according to face detection through the generation of a look-up table accurately standardized and perform a real-time operation. A census transformation method is applied to an input image, and a feature value is extracted. Face detection learning is performed by using the extracted feature value, and a look-up table having a reference feature value group is obtained. Real face detection is performed by using the feature value of the obtained look-up table.
摘要翻译：提供了一种在照明变化中较强的面部检测方法，以通过产生准确地标准化的查找表来减少根据人脸检测的计算量并执行实时操作。将普查变换方法应用于输入图像，并且提取特征值。通过使用提取出的特征量进行面部检测学习，得到具有基准特征量组的查询表。通过使用获得的查找表的特征值来执行真实的脸部检测。

3. 发明公开

KR1020080079798A 얼굴 검출 및 인식을 위한 방법 无效
标题翻译：面部检测和识别方法
公开(公告)号：KR1020080079798A
公开(公告)日：2008-09-02
申请号：KR1020070020251
申请日：2007-02-28
申请人： (주)코아정보시스템
发明人： 강태운
IPC分类号： G06K9/46 , G06T7/00 , G06K9/80
CPC分类号： G06K9/00275 , G06K9/4614 , G06K9/6256 , G06T5/009
摘要： A method for detecting and recognizing a face is provided to recognize the face correctly with a small calculation quantity through face detecting and appearance-based approaching techniques robust to lighting change regardless of an image size by detecting the face based on a feature-based approaching technique. A face image is captured and filtered to remove noise(200). The filtered face image is reduced to a predetermined window size(300). A local structure feature of a face is extracted by applying MCT(Modified Census Transform) to the reduced face image(400). A candidate face image is selected by comparing the extracted local structure feature of the face with a predetermined threshold value(500). A representative area suitable for the data is selected among the candidate face images. The face is finally detected by performing previous stages with the selected representative area(700). The face is recognized by using the finally detected face.
摘要翻译：提供了一种用于检测和识别面部的方法，以通过面部检测和基于外观的接近技术，以小的计算量正确地识别面部，而不管图像尺寸如何，通过基于特征的接近技术检测脸部而不考虑图像尺寸。拍摄和过滤脸部图像以消除噪音（200）。滤波后的脸部图像被缩小到预定的窗口尺寸（300）。通过将MCT（修改人口普查变换）应用于缩小面部图像（400）来提取脸部的局部结构特征。通过将所提取的面部局部结构特征与预定阈值（500）进行比较来选择候选面部图像。在候选面部图像中选择适合数据的代表区域。通过使用所选择的代表区域（700）执行以前的阶段，最终检测出脸部。通过使用最终检测到的脸部识别脸部。

4. 发明公开

KR1020080101388A 엔엠씨티 기반의 얼굴 검출 방법 无效
标题翻译：基于新的修正变量的面部检测算法
公开(公告)号：KR1020080101388A
公开(公告)日：2008-11-21
申请号：KR1020070048081
申请日：2007-05-17
申请人： (주)코아정보시스템
发明人： 강태운 , 류무호 , 이수현 , 김영진
IPC分类号： G06K9/46 , G06T7/40 , G06K9/36
CPC分类号： G06K9/00248 , G06K9/00288 , G06K9/42
摘要： A face detecting method based on MCT(New Modified Census Transform) is provided to reduce the waste of the memory by controlling the storage bit number and use the memory efficiently by transforming an image in transforming CT(Census Transform) as a new MCT(New Modified Census Transform). A resized image is transformed in new MCT by using the position feature of image(S102). Face detecting window is applied to the resized image by obtaining coordinates of each filter table. A final result is outputted by grouping face candidates regions of face detected by various sizes and positions(S104). The face candidate region is selected by calculating the size of the represent face candidate and a weight mean of center pixel coordinates(S105).
摘要翻译：提供了基于MCT（新修改的普查变换）的面部检测方法，通过控制存储位数减少存储器的浪费，并通过在转换CT（Census Transform）中转换图像作为新的MCT（New 修改人口普查变换）。通过使用图像的位置特征（S102），在新的MCT中变换调整大小的图像。通过获得每个滤波器表的坐标，将面部检测窗口应用于调整大小的图像。通过对由各种尺寸和位置检测到的面部候选区域进行分组来输出最终结果（S104）。通过计算代表面部候选者的大小和中心像素坐标的权重平均值来选择面部候选区域（S105）。

你已经成功收藏专利！

检索式保存成功!

IPRDB

热门服务

关于我们

友情链接

联系方式