用多消失矩最优小波包基改进虹膜识别

Improving Iris Recognition by Multi-vanishing Moments Joint Best Basis Selection

采用层叠算法求出尺度函数和小波函数的离散采样序列的逼进,将不具有基函数解析表达式的母小波用于虹膜特征提取,并利于实现连续小波包变换.针对虹膜光学识别应用背景,提出采用基于连续小波包变换的多消失矩联合最优小波包基来改进特征图像相关识别的最优基,并用统计识别方法进行后处理以增强算法适应性,同时提出通过体全息相关系统来实现,以发挥光学高并行性的优势,模拟结果表明可获得比已有方法更高的识别率.

In digital computation, it is the discrete sampling sequences of continuous wavelet functions that are used. Using the cascade algorithm, the discrete approximating sequences of scaling and wavelet functions are computed for introducing the mother wavelets without analytical forms into iris feature extraction. By the definition of wavelet packets, the discrete approximating sequences of wavelet packet functions can also be computed to fulfill continuous transform. Using continuous wavelet packet transform based on these sequences, the multi-vanishing moments joint best wavelet packet bases are chosen for eigen-images generation in the eigen-images based correlation recognition. This recognition is implemented by a volume holographic correlation system to take good use of high parallelism of optics. The modified post-processing method using statistic feature can make the algorithm more robust to the errors introduced in optical system. In simulation, with the high precision of digital computation, the identification rate is 90.91% and is higher than the identification rate, 85. 2%, which is obtained by the dual multi-channel statistic recognition method using the same mother wavelet, Db4.