【摘要】 以原始人脸图像为基础,用灰度拉伸、灰度均衡、中值滤波、图像锐化和图像变换等多种处理方法对人脸图像进行预处理,用分形几何的关联维方法提取人脸图像的特征值,并通过BP网络建立性别识别模型。在此基础上,用上述各种预处理方法对29个未参与建模的人脸图像处理,然后进行性别识别。结果表明,用灰度拉伸和傅立叶变换处理的人脸图像,性别识别效果较好;图像变换比其他处理方法的识别效果好;先对人脸图像作灰度拉伸,然后作傅立叶变换,性别识别效果最佳。更多还原

【Abstract】 On the basis of primitive face images,the face images are preprocessed with gray stretching,gray equalizing,median filtering,image sharpness and image transforms,while the character values of face images are abstracted through the correction dimension algorithm in fractals and the gender recognition models are made with the BP network.And then 29 face images without being modeled are processed with the above preprocessing methods before the gender recognition is taken.From the processed data,it shows that the gender recognitions through the gray stretching and Fourier transform of face images are better,and the gender recognitions of image transforms are better than that of other processing methods,and the gender recognition is best through the combination of gray stretching and Fourier transform of face images.更多还原