This paper develops an efficient classification algorithm called UDP, which reduces the high dimension of sample to low dimensional subspace simply said as dimensionality reduction.UDP takes into account both the local and non-local quantities. It can well characterize the local scatter as well non-local scatter which simultaneously maximizes the non–local scatter and minimizes the local scatter. This makes UDP more powerful and more intuitive than LDA and PCA. This makes UDP a good choice for real-world biometrics application The proposed method is applied to palm biometrics and is examined by small set of samples per class.

-This paper develops an efficient classification algorithm called UDP, which reduces the high dimension of sample to low dimensional subspace simply said as dimensionality reduction.UDP takes into account both the local and non-local quantities. It can well characterize the local scatter as well non-local scatter which simultaneously maximizes the non–local scatter and minimizes the local scatter. This makes UDP more powerful and more intuitive than LDA and PCA. This makes UDP a good choice for real-world biometrics application The proposed method is applied to palm biometrics and is examined by small set of samples per class.