基于离散余弦变换和主线分块能量的模糊掌纹识别

Blurred palmprint recognition based on DCT and block energy of principal lines

针对非接触式掌纹采集时离焦状态导致的图像模糊问题,提出一种新颖的识别方法。使用离散余弦变换(DCT)在频域内提取低频系数作为稳定特征,使用改进的局部灰度极小值法提取空域内的稳定特征即主线,再使用分块方法计算主线能量形成特征向量,然后将频域和空域内的稳定特征进行融合,最后利用向量之间的欧式距离进行识别。在SUT-D模糊掌纹库上的测试结果表明,与融合之前及其他典型识别方法比较,本文算法识别率最高可达96.057 8%,表明本文方法在识别性能上具备有效性和优越性,为解决模糊掌纹的识别问题提供了一条可行途径。

In view of the problem of blurred image caused by defocus status for non-contact palmprint col lection, a novel recognition approach is proposed. As the stable features, the low frequency coefficients are extracted by discrete cosine transform（DCT） in the frequency domain, and the principal lines are extrac- ted by the improved local gray minimum method in the spatial domain. The block method is used for cal culating principal lines energy to form the feature vectors, then the stable features in the frequency and spatial domains are {used, and finally the Euclidean distance between vectors is used for classification and identification. The experiments based on the SUT-D blurred palmprint database show that compared with no-fusion and other typical identification methods, the proposed algorithm can get recognition rate up to 96. 057 8%, which means that it is an effective and superior approach to solve the problem of blurred palmprint recognition.