Compressive imaging is a new imaging method based on the compressive sensing theory, which has the advantage of sparse/compressible image reconstruction with far fewer measurements than traditional Nyquist samples. By analyzing the existing Bernoulli and Toeplitz matrix, we propose a novel sparse trinary Toeplitz matrix with random pitch for phase mask. Simulation results show that novel phase mask matrices, comparing with Bernoulli and Bernoulli-Toeplitz phase mask matrices, almost have the same signal-to-noise ratio. But with dramatical reduction of the number of independent random variables and the number of non-zeros entries, the novel matrix is more conducive to data transmission and storage and easy for hardware implementation. Even more, the reconstruction time is only about 21%~66% to that of original matrices.