In practical applications of medical imaging systems,images with high resolution are always needed.However,this will lead to large data that makes it difficult to store a nd transfer.Motivated by the idea of compressed sensing,a nonlinear sampling theory can be used to reconstru ct signal from incomplete frequency data.This paper tries to find out a better sampling me thod to gain high resolution images as well as reduce the data storage.Based on the sparsity or compressibil ity,signals can be sampled by the rate lower than Nyquist sampling rate.Compressed sensing is appli ed in medical image compressive sampling and a new method called partitioning orthogonal matching pursuit with dual threshold s is proposed in this paper.Due to the difference in blocks information quantity,the information quant it y is estimated by sampling threshold to select different sampling rates for different blocks,which can imp rove the sampling and reconstruction efficiency.In order to obtain good reconstruction quality,a jud ging threshold is used to reduce the dependence on the sampling threshold.The experimental results show th at this method can reconstruct an image with higher accuracy and lower samplin g rate.