Traditional correction methods can not be used to correct effectively the solar images with uneven illumination, such as the bilinear interpolation method. We adopt an improved algorithm that combine the background fitting method and the mask method. The algorithm consists of the following main steps: segmenting the image, selecting the sampling points, interpolating the sampling points, calculating the mask, correcting the image. By comparing four evaluation indicators of the corrected image, including the information entropy of the image, the mean brightness, the mean variance and the peak signal-to-noise ratio (PSNR), this improved algorithm is proved to be effectively in the solar images with uneven illumination.