CAJ | 학술논문

针对单演信号表述仅利用幅值和方向而忽略相位信息的问题，提出了一种单演同相幅值模式方法。用多尺度的单演滤波器提取人脸的单演幅值和相位信息，将相位量化并根据相位量化的结果对幅值累加和二值编码，从而获得若干张单演同相幅值模式图。将每一张PMMSP图分块并提取直方图特征，用BFLD对特征进行降维。这样能有效提升特征的判别能力并降低算法的时间和空间复杂度。在CAS-PEAL人脸库和AR人脸库上的实验证明了该算法的有效性。
침대단연신호표술부이용폭치화방향이홀략상위신식적문제，제출료일충단연동상폭치모식방법。용다척도적단연려파기제취인검적단연폭치화상위신식，장상위양화병근거상위양화적결과대폭치루가화이치편마，종이획득약간장단연동상폭치모식도。장매일장PMMSP도분괴병제취직방도특정，용BFLD대특정진행강유。저양능유효제승특정적판별능력병강저산법적시간화공간복잡도。재CAS-PEAL인검고화AR인검고상적실험증명료해산법적유효성。
In order to take advantages of the monogenic magnitude and phase information for face recognition, a new method based on pattern of monogenic magnitudes with same phase is proposed. Multi-scale monogenic filters are used to extract the monogenic magnitude and phase information of the image. This paper obtains the phase quantization images, accumu-lates the magnitude according to the quantization results and codes the accumulative results by binary strategy. After that the patterns of monogenic magnitudes with same phase image are obtained. PMMSP maps are divided into blocks and the histo-gram is extracted from the blocks. BFLD is not only used to reduce the dimension of the features but also enhance its discrimi-native power. The experimental results on the CAS-PEAL and AR face databases show that the proposed algorithm is effective.