Massive multi-input multi-output (MIMO) has been regarded as one of the key technologies for fifth generation (5G) mobile communication systems, as it can significantly enhance the system capacity with high spectrum and energy efficiency. For massive MIMO systems, accurate channel estimation is a challenging problem, especially when the number of parameters to be estimated is large and the number of pilots is limited. In this paper, a compression-based linear minimum mean square error (CLMMSE) channel estimation algorithm is proposed for massive MIMO in 5G systems. Compared with the traditional linear minimum mean square error (LMMSE) algorithm, the proposed approach calculates the channel autocorrelation matrix by investigating the channel prior information based on compressive sensing (CS) theory, utilizing the block sparsity of massive MIMO channels, to reduce the complexity for obtaining autocorrelation matrix. Then it substitutes matrix inverse operation by singular value decomposition to further reduce the computational complexity. In addition, a block sparsity adaptive matching pursuit (BSAMP) method is also proposed to adaptively estimate the block sparsity of the channel in the first step of the proposed CLMMSE algorithm, which can make it more efficient. The sparsity-adaptive processing is achieved by setting a threshold and finding the position of the maximum backward difference, then using the regularized method to solve channel estimation as a convex optimization problem. Analyses and simulations indicate that with slight performance degradation, the proposed algorithm reduces the computational complexity significantly compared with the traditional LMMSE algorithm. And compared with pure CS methods, CLMMSE has an obviously better performance, which is beneficial to solve the pilot pollution problem of massive MIMO in 5G systems. Furthermore, the BSAMP-based CLMMSE algorithm has better performance and lower time complexity than the algorithm based on other CS methods, which further improves the system performance.