Biometric identification was a kind of identity recognition technology by making use of the human's unique physiological or behavioral characteristics,it provided a high reliability and stability way for the ident...Biometric identification was a kind of identity recognition technology by making use of the human's unique physiological or behavioral characteristics,it provided a high reliability and stability way for the identification. Global threshold binarization palmprint image is used in this paper,and the bio-morphological methods are used to get the sensitive area of palmprint image's positioning point,so as to extract the region of interest. The palmprint collection is realized on the FPGA chip,and this kind of collection method uses the DSP Builder toolbox to realize visual programming in Matlab / Simulink and achieve fast modeling and development. The practice proves that this method is simple,flexible and its equipment is portable and fast.展开更多
感兴趣区域(Region of interest,ROI)提取在生物特征识别中,常用于减少后续处理的计算消耗,提高识别模型的准确性,是生物识别系统中预处理的关键步骤.针对生物识别数据,提出了一种鲁棒的ROI提取方法.方法使用语义分割模型作为基础,通过...感兴趣区域(Region of interest,ROI)提取在生物特征识别中,常用于减少后续处理的计算消耗,提高识别模型的准确性,是生物识别系统中预处理的关键步骤.针对生物识别数据,提出了一种鲁棒的ROI提取方法.方法使用语义分割模型作为基础,通过增加全局感知模块,与分割模型形成对抗结构,为模型提供先验知识,补充全局视觉模式信息,解决了语义分割模型的末端收敛困难问题,提高了模型的鲁棒性和泛化能力.在传统二维(2D)指纹、人脸、三维(3D)指纹和指纹汗孔数据集中验证了方法的有效性.实验结果表明,相比于现有方法,所提出的ROI提取方法更具鲁棒性和泛化能力,精度最高.展开更多
In this paper, more efficient, low-complexity and reliable region of interest （ROI） image codec for compressing smooth low texture remote sensing images is proposed. We explore the efficiency of the modified RO! cod...In this paper, more efficient, low-complexity and reliable region of interest （ROI） image codec for compressing smooth low texture remote sensing images is proposed. We explore the efficiency of the modified RO! codec with respect to the selected set of convenient wavelet filters, which is a novel method. Such ROI coding experiment analysis representing low bit rate lossy to high quality lossless reconstruction with timing analysis is useful for improving remote sensing ground truth surveillance efficiency in terms of time and quality. The subjective [i.e. fair, five observer （HVS） evaluations using enhanced 3D picture view Hyper memory display technology] and the objective results revealed that for faster ground truth ROI coding applications, the Symlet-4 adaptation performs better than Biorthogonal 4.4 and Biorthogonal 6.8. However, the discrete Meyer wavelet adaptation is the best solution for delayed ROI image reconstructions.展开更多
针对JPEG2000中隐式ROI(Region of Interest)编码算法在低码率下重建图像感兴趣区质量较差的问题,提出了一种改进方法.利用拉格朗日插值法构造权重自适应函数,根据ROI码块中ROI小波系数个数所占比例及设定的目标码率,合理地为ROI码块分...针对JPEG2000中隐式ROI(Region of Interest)编码算法在低码率下重建图像感兴趣区质量较差的问题,提出了一种改进方法.利用拉格朗日插值法构造权重自适应函数,根据ROI码块中ROI小波系数个数所占比例及设定的目标码率,合理地为ROI码块分配权重,在保证更多ROI信息被优先编码的同时,降低背景小波系数的影响,从而提高了ROI小波系数在码流中的数据长度.实验结果表明,采用该算法,重建后感兴趣区图像质量较隐式ROI算法有明显提高.展开更多
文摘Biometric identification was a kind of identity recognition technology by making use of the human's unique physiological or behavioral characteristics,it provided a high reliability and stability way for the identification. Global threshold binarization palmprint image is used in this paper,and the bio-morphological methods are used to get the sensitive area of palmprint image's positioning point,so as to extract the region of interest. The palmprint collection is realized on the FPGA chip,and this kind of collection method uses the DSP Builder toolbox to realize visual programming in Matlab / Simulink and achieve fast modeling and development. The practice proves that this method is simple,flexible and its equipment is portable and fast.
文摘感兴趣区域(Region of interest,ROI)提取在生物特征识别中,常用于减少后续处理的计算消耗,提高识别模型的准确性,是生物识别系统中预处理的关键步骤.针对生物识别数据,提出了一种鲁棒的ROI提取方法.方法使用语义分割模型作为基础,通过增加全局感知模块,与分割模型形成对抗结构,为模型提供先验知识,补充全局视觉模式信息,解决了语义分割模型的末端收敛困难问题,提高了模型的鲁棒性和泛化能力.在传统二维(2D)指纹、人脸、三维(3D)指纹和指纹汗孔数据集中验证了方法的有效性.实验结果表明,相比于现有方法,所提出的ROI提取方法更具鲁棒性和泛化能力,精度最高.
基金Project (No. 2004144013) supported by the Chinese Government Scholarship Council, China
文摘In this paper, more efficient, low-complexity and reliable region of interest （ROI） image codec for compressing smooth low texture remote sensing images is proposed. We explore the efficiency of the modified RO! codec with respect to the selected set of convenient wavelet filters, which is a novel method. Such ROI coding experiment analysis representing low bit rate lossy to high quality lossless reconstruction with timing analysis is useful for improving remote sensing ground truth surveillance efficiency in terms of time and quality. The subjective [i.e. fair, five observer （HVS） evaluations using enhanced 3D picture view Hyper memory display technology] and the objective results revealed that for faster ground truth ROI coding applications, the Symlet-4 adaptation performs better than Biorthogonal 4.4 and Biorthogonal 6.8. However, the discrete Meyer wavelet adaptation is the best solution for delayed ROI image reconstructions.
文摘针对JPEG2000中隐式ROI(Region of Interest)编码算法在低码率下重建图像感兴趣区质量较差的问题,提出了一种改进方法.利用拉格朗日插值法构造权重自适应函数,根据ROI码块中ROI小波系数个数所占比例及设定的目标码率,合理地为ROI码块分配权重,在保证更多ROI信息被优先编码的同时,降低背景小波系数的影响,从而提高了ROI小波系数在码流中的数据长度.实验结果表明,采用该算法,重建后感兴趣区图像质量较隐式ROI算法有明显提高.
