针对照明变化条件下人脸图像检测精度相对较低的问题,以照明变化下的人脸检测为研究对象,提出局部自我相关函数(local autocorrelation,LAC),研究基于Adaboost算法下采用局部自我相关函数为前处理的光照变化下人脸检测。提出了局部自我...针对照明变化条件下人脸图像检测精度相对较低的问题,以照明变化下的人脸检测为研究对象,提出局部自我相关函数(local autocorrelation,LAC),研究基于Adaboost算法下采用局部自我相关函数为前处理的光照变化下人脸检测。提出了局部自我相关函数定义模型,对局部自我相关函数的物理特性进行分析,从理论上验证局部自我相关函数对线性照明变化的鲁棒性。采用卡内基梅隆大学的人脸照明变化数据库(CMU PIE Database)作为检测数据验证基于局部自我相关函数的光线照明变化下的人脸检测,实验结果证明了局部自我相关函数消除照明变化对人脸检测精度影响的有效性。展开更多
Early detection of Non-Proliferative Diabetic Retinopathy (NDPR) is currently a highly interested research area in biomedical imaging. Ophthalmologists discover NDPR by observing the configuration of the vessel vascul...Early detection of Non-Proliferative Diabetic Retinopathy (NDPR) is currently a highly interested research area in biomedical imaging. Ophthalmologists discover NDPR by observing the configuration of the vessel vascular network deliberately. Therefore, a computerized automatic system for the segmentation of vessel system will be an assist for ophthalmologists in order to detect an early stage of retinopathy. In this research, region based retinal vascular segmentation approach is suggested. In the steps of processing, the illumination variation of the fundus image is adjusted by using the point operators. Then, the edge features of the vessels are enhanced by applying the Gabor Filter. Finally, the region growing method with automatic seed point selection is used to extract the vessel network from the image background. The experiments of the proposed algorithm are conducted on DRIVE dataset, which is an open access dataset. Results obtain an accuracy of 94.9% over the dataset that has been used.展开更多
文摘针对照明变化条件下人脸图像检测精度相对较低的问题,以照明变化下的人脸检测为研究对象,提出局部自我相关函数(local autocorrelation,LAC),研究基于Adaboost算法下采用局部自我相关函数为前处理的光照变化下人脸检测。提出了局部自我相关函数定义模型,对局部自我相关函数的物理特性进行分析,从理论上验证局部自我相关函数对线性照明变化的鲁棒性。采用卡内基梅隆大学的人脸照明变化数据库(CMU PIE Database)作为检测数据验证基于局部自我相关函数的光线照明变化下的人脸检测,实验结果证明了局部自我相关函数消除照明变化对人脸检测精度影响的有效性。
文摘Early detection of Non-Proliferative Diabetic Retinopathy (NDPR) is currently a highly interested research area in biomedical imaging. Ophthalmologists discover NDPR by observing the configuration of the vessel vascular network deliberately. Therefore, a computerized automatic system for the segmentation of vessel system will be an assist for ophthalmologists in order to detect an early stage of retinopathy. In this research, region based retinal vascular segmentation approach is suggested. In the steps of processing, the illumination variation of the fundus image is adjusted by using the point operators. Then, the edge features of the vessels are enhanced by applying the Gabor Filter. Finally, the region growing method with automatic seed point selection is used to extract the vessel network from the image background. The experiments of the proposed algorithm are conducted on DRIVE dataset, which is an open access dataset. Results obtain an accuracy of 94.9% over the dataset that has been used.
