Abstract: In consideration of the the issue of the weakness performance of Compressive Sensing (CS) ISAR imaging under low SNR condition, an improved CS method is proposed in this paper. Energy based threshold is proposed to identify range cells containing only noise, a coherent projection in cross-range to improve the SNR of measurements then a iterative re-weighting norm optimization is applied to enhance the signal recovery while reject noise. Moreover, an advantage of this robustness is the ability to produce a high quality image and is suitable for ISAR imaging by using very limited echoes under strong noise and clutter. Experimental result of real data processing verifies the proposed methods capability of depressing noise or clutter and extracting strong scatterers to construct high resolution image. In the experiment, high quality image can be generated by using only 16 pulses and its robustness is confirmed too.