In order to correct the non-uniformity (NU) of a single infrared (IR) image and optimize the result of details,a locally adaptive non-uniformity correction algorithm is proposed. According to the distribution character of the fixed pattern noise in infrared focal plane array,median histogram equalization non-uniformity algorithm based on Gaussian weights is used to realize single parameter correction of infrared image. Then all images are decomposed into patches,and each patch of images can adaptively select the most suitable correction parameter,and can realize detail optimization. The experiment results show that the proposed algorithm in root-mean-square error(RMSE),peak signal to noise ratio (PSNR) and the smooth of image is improved further compared with single parameter median histogram equalization non-uniformity correction algorithm,and it reserves more details of images,which provides a new method for non-uniformity correction.