The invention discloses a face recognition method based on the optimized kernel Fukunaga-Koontz transformation. The method comprises the steps of extracting the image features out of a training sample, selecting a Gaussian kernel function, optimizing the Gaussian kernel function, conducing the data dimension reduction of the sample, calculating the autocorrelation matrix and the whitening matrix of each sample, determining the category of each face, and the like. According to the technical scheme of the invention, the kernel optimization is realized based on the KFKT method, so that the recognition performance is improved.