CAJ | 학술논문

复杂光照条件会引起人脸表面的光照分布不均匀,出现局部过亮、过暗和反光等问题,给后续人眼定位带来较大影响.为了提高复杂光照条件下人眼定位的准确率,提出一种先定位人脸再定位人眼的方法,该方法首先采用高斯肤色模型结合人脸先验知识定位人脸区域,对所得人脸区域进行自适应连续均值量化变换增强,最后结合连通域分析对人眼进行定位.实验结果表明,该方法对复杂识别率较高,对头部姿势及面部表情等的变化具有一定的鲁棒性.
복잡광조조건회인기인검표면적광조분포불균균,출현국부과량、과암화반광등문제,급후속인안정위대래교대영향.위료제고복잡광조조건하인안정위적준학솔,제출일충선정위인검재정위인안적방법,해방법수선채용고사부색모형결합인검선험지식정위인검구역,대소득인검구역진행자괄응련속균치양화변환증강,최후결합련통역분석대인안진행정위.실험결과표명,해방법대복잡식별솔교고,대두부자세급면부표정등적변화구유일정적로봉성.
The complex illumination conditions can cause a problem of uneven distribution of light on face surface which makes parts of face image bright, dark or reflective and bring great influence on subsequent eye location.To increase the accuracy of eye location under complex illumination, a method which locates face before eyes is proposed. Gaussian skin model is built to locate face region with the help of prior knowledge.And a self-adaptive successive mean quantization transform image enhancement algorithm is used to enhance the face region. Eyes are located by analyzing the connected component. The results showed that this method has a high rate of eye detection under complex illumination conditions and it is robust to complex posture and variant expression.