Abstract: To improve the disadvantages that iterative reconstruction algorithms of compressed sensing need priori knowledge of the sparsity of original signal or iterative threshold,an adaptive sparse recovery based on difference algorithm is proposed.When the sparsity of original signal is unknown,the proposed algorithm takes advantage of unbalance of correlation coefficient between the measurement matrix and residual.With those properties,the proposed algorithm can select the support set of the original signal,and approach the sparsity of the original signal.Simulation results show that the proposed algorithm obtains better recovery results under the same conditions.Especially in the lower sampling rate,the advantage is more obvious.