关键词: 图像纹理特征, 二维非高斯Gabor滤波器, 生物图像表示, 人脸和掌纹识别

Abstract: Gabor filter has gained much reputation as a powerful and attractive texture operator in terms of accuracy. However, Gabor filter can be viewed as a Gaussian kernel modulated by a complex sinusoid in the frequency domain. Important information of images contained in the frequency bands lying outside the coverage of the Gaussian kernel bandwidth can not be extracted by Gabor filters. This paper develops another non-Gaussian Gabor filter with a new introduced parameter for texture feature extraction. The newly introduced parameter can control the shape of envelope of non-Gaussian Gabor filters. To demonstrate the superiority of the proposed non-Gaussian Gabor filter, extensive experiments on biometrics have been done. Experimental results have shown that the performance of the proposed non-Gaussian Gabor filter has larger superiority than Gabor filter.

Key words: image texture feature, non-Gaussian Gabor filter, biometric image representation, face and palmprint recognition