Stripe noise perturbs image qualities of push-broom-type aerial camera, thus eliminating stripe noise is of vital importance to improve the precision of posterior data analysis. The main sources and the model of stripe noise in aerial images are analyzed and a new destriping stripe noise algorithm based on total variation is proposed. The gain and offset of each pixel are estimated based on quasi-homogeneous regions. The total variation algorithm is utilized and the image is reconstructed by the gradient descent method iteratively. The qualitative analysis results demonstrate that the peak signal-to-noise ratio of the simulated image is improved from 31 dB to 40 dB and the improvement factor of radiometric quality of the real aerial image is improved to 9 dB. Compared with traditional methods, the proposed algorithm can achieve higher inverse coefficient of variation and improvement factors of radiometric quality. Stripe noise can be effectively removed with detailed information of the original image reserved.