Aiming at the problem that the existing far-field and near-field channel estimation schemes would not be directly used to accurately estimate the mixed-field channel state information, an effective mixed-field channel estimation scheme based on Compression Sampling Matching Pursuit (CoSaMP) was proposed. The hybrid-field channel model was accurately modeled to capture the feature that different scatterers may exist in the far-field and near-field regions, and the far-field and near-field path components were estimated respectively. The results show that compared with other channel estimation schemes, the proposed scheme can obtain better normalized mean square error performance and lower computational complexity under the same low pilot overhead. It is concluded that the proposed scheme has good reference value.