The invention discloses a non-uniformity correction method based on fuzzy domain singular value decomposition (SVD). Firstly, an image is converted to a fuzzy domain from a spatial domain through a membership function, and the fundamental frequency information of the image is obtained through SVD in the fuzzy domain; then, the image after processing is converted to the spatial domain from the fuzzy domain, subtraction with an original image is carried out to remove stripe noise and lid effects in the image; and finally, a noiseless image after correction is obtained through a stretching contrast algorithm. The non-uniformity correction method based on the SVD has the advantages that the fuzzy domain method is firstly introduced for non-uniformity correction on the infrared image, the fundamental frequency information of the image is obtained through the SVD, the information is non-uniformity noise extracted from the image, the whole processing process only needs to use single frame information, drastic requirements on a scene by the traditional non-uniformity correction algorithm are avoided, and the method can be widely applied to non-uniformity correction tasks in different environments.