动态交叉模板与方向滤波融合的掌纹褶皱检测方法.pdfVIP

万方数据 ABSTRACT The recognition accuracy of existing palmprint identification systems greatly depends on the quality of the palmprint image. There are many crease regions in the palmprint image. These crease regions, which have high gray value, will cut the ridge and valley off and form some special patterns which are different from the valley. Detecting crease regions in the palmprint image is very meaningful. Firstly, the crease regions will affect the enhancement of palmprint seriously and in the crease regions there will be many false minutiae which will greatly affect the recognition accuracy. Second, the crease in the palmprint image is a special pattern for identification. This thesis propose a novel method that fusion the method based on dynamic cross mask and orientation filter for palmprint crease detection. Based on the special pattern of the crease, a new algorithm is proposed. This thesis focuses on three aspects. First, based on the special pattern of the crease, this thesis propose a dynamic cross mask method, and the method of constructing the crease cross mask and enhanced mask is proposed. Second, a new method named orientation filtering method, which is a improved second-order Gaussian filtering method, is proposed, and the coherence of the six directions of a image block is analyzed to get several latent directions of crease. And the orientation filtering method also can be used to detecting crease in palmprint. At last, a me