The compressive sensing reconstruction of magnified in-line hologram was experimentally researched in this paper. It aimed at the tomographic reconstruction of magnified multilayer samples. Firstly, the compressive sensing reconstruction of in-line hologram was introduced in theory, and the realization procedure was shown, including frequency domain down-sampling mode, the flow of two-step iterative algorithms, etc. Secondly, hologram recording system in case of point-source magnification and microscope objective magnification was established, the experimental work was carried out, which took multilayer samples as an example, and tomographic reconstruction of recorded in-line hologram based on compressive sensing. In the same time, it processed back-propagation reconstruction based on traditional convolution algorithm. The experimental results show that compressive sensing tomographic reconstruction technology can realize the tomographic reconstruction of holograms obtained in two magnification modes, and have better reconstruction results than the traditional convolutional inverse diffraction reconstruction. And it shows the ability and superiority of compressive-sensing tomographic reconstruction.