In order to combine the advantages of global threshold with local threshold method, a double-threshold binarization method based on image enhancement was proposed. The two thresholds were calculated out by using OTSU and Kittler method, and preprocessing was made to the points with gray value greater than the upper threshold or less than the lower threshold. Sauvola method was used to judge the points with gray value between the two thresholds. Double-threshold remained more detailed information with less time cost, but still had the problem of uneven illumination. Therefore, a new image enhancement algorithm was proposed. The image was divided into four areas equally, and the differences of the average grey value of the whole image with that of each area were taken. Then by using the four weights related to pixel points'location, the corresponding gray value on each point was compensated. Image enhancement was made repeatedly until the average gray value of the four areas was controlled within a certain range. The binarization of uneven illumination image was improved greatly after this image enhancement.