Compressive Sensing (CS) indicates that when the signal of interest is sparse or compressible (i.e., sparse after mathematical transformation), one can take a small number of linear projection measurements from the signal, and reconstruct the signal almost perfectly through proper algorithm. The feature of the CS has great potential applications in that high-resolution imaging is highly desirable while large size detector array is unavailable, such as those in ultraviolet or infrared wavelength region or that in aircrafts and satellites working condition when the data transmission is a key issue.<p> </p>However, CS technique still faces challenges in the signal sampling and reconstruction. Firstly, detector measurements must be nonnegative in linear optical system which is different from digital image processing. Secondly, blurring caused by practical optical system should be considered, which will destroy the effect of reconstruction.<p> </p>In this paper, we discuss some kinds of phase encoding which could be used in practice imaging system. We make a compensation to solve the non-negative problem when CS applied in the practical optical system, use a small size detector to receive a general image degrading model, and reconstructed image from the single, low-solution and noisy observation through a fast and feasible non-linear algorithm, the result proves our system is robust and feasible.