研究了一种基于灰度数学形态学和二维小波变换的指节折痕特征提取和识别算法。该算法首先利用灰度形态学对指节折痕子图进行预处理增强,通过选择合适的结构元素来施行对指节折痕子图图像进行灰度形态学处理,大大地改善了折痕线的对比度,然后利用二维小波分解的方法将增强后的图像进行二维分解,得到不同分辨率和不同方向的小波变换子图,最后根据折痕线的分布特征选择所需子图并计算它们的不变矩构成特征矢量来表征指节折痕特征。实验结果验证了方法的有效性。
A new approach to finger crease feature extraction by using mathematical morphology and 2-D wavelet transform are studied. Firstly, morphology operations are used to enhance finger crease sub-image in different directions, since the shape of the structuring element affects the result of crease line enhancement, the different directional structuring element is used to enhance the crease line. Secondly, the enhanced sub-image is decomposed in several directions and at different resolutions with 2-D wavelet transformation to obtain wavelet sub-images. Finally, invariant moments of wavelet sub-images are used to form feature vector. Experimental results illustrate the effectiveness of our method.