This article establishes some theoretical results about the maximum function of probability density functions ( fmax) and the integration of fmax (I fmax). Using the probability density function extracted from the image as a relatively stable feature of the image and I fmax as a measure the similarity between a “face” candidate region and a group of training face images, we propose a new face detection method, one of the most challenging tasks related to image analysis. The experiments demonstrate the competitiveness of the proposed method, especially in the case of rotated images. It also shows potential in real application of the researched problem.