An optical distortion-invariant image recognition method based on the multivariated statistical analysis is presented. In this approach, a set of eigenimages is first extracted from a large number of training images including various distortions by using the principal component analysis and then are used as the reference patterns to be optical correlated with the testing input image. The optical correlation results between the input image and the set of eigenimges construct a feature space, on which the discriminant analysis is performed during the training and classification process. Then the distortion-invariant recognition to the input image can be implemented quickly. The optical experimental results implemented on an incoherent optical correlator are given.