Satellite vibration and scan mirror oscillation can cause oscillation distortions of remote sensing imagery. These distortions are difficult to be corrected by common geometric correction methods. A geometric correction method using compressive sampling is proposed. In this method, geometric correction is performed with rational function model (RFM). The biases between the image coordinates and project coordinates of ground control points (GCPs) are calculated with the original RFM. The bias is called projection bias. The projection biases at the GCPs are regarded as measurements. By using the technique of compressive sampling, the projection biases of every pixel can be reconstructed, and the RFM is compensated according to projection biases. The remote sensing imagery is rectified with the compensated RFM. Through compensation, the errors of RFM which caused by oscillation distortions are eliminated and the performance of geometric correction is improved. The method is proved effective by using real remote sensing images. Based on simulated images, the effect of GCPs distribution and number on geometric correction is also analyzed.