For the poor effect of non-uniform illumination image and long time computing by the traditional local threshold segmentation algorithm, an image binarization optimization algorithm based on local mean is proposed. The algorithm uses the fast integral image algorithm to calculate the local mean and then substitutes the local mean into the improved model of the threshold algorithm to complete segmentation, thereof improves the image segmentation effect in non-uniform illumination. The contrast experiment results indicate that the proposed algorithm greatly improves the computation efficiency and enhances the image segmentation effect, which is superior to Niblack algorithm, Sauvola algorithm, Bernsen algorithm as well as local contrast and mean algorithm both in visual effect and processing time.