Aiming to resolve the classification problem that image samples are multi-subclass distributed and non-linearly separable

a kernel two-dimensional subclass discriminant analysis algorithm (K2DSDA) is proposed.In this paper

it has shown that K2DSDA algorithm is theoretically equivalent to column/row-2DSDA based algorithm on kernel samples.Meanwhile

the optimal discriminant vectors are computed via the approximate kernel samples

so that the computational complexity is greatly reduced.The experimental results which tested on benchmark face database show that the proposed algorithm is superior to other state-of-the-art discriminant analysis algorithms

which confirms the effectiveness of K2DSDA.