Although infrared focal plane imaging technology has been widely used in target detection and orientation, reducing pixel-pitch or increasing the number of infrared sensor array pixels was unable to meet the practical requirement for the restriction of infrared sensor’s processing technology. This paper proposes a approach that apply multi-element detector array in CS imaging system. The coded measurement technology which uses optical mask and realizes compressed sampling of target scene is described in this paper, on the basis of summarizing CS theory and algorithm. Considering the fact that the reconstruction process of large size image lasts a long time, this paper also presents a high quality image partition CS method, which requires less computing time. The simulation results point the best choice for image partition which requires the least computing time. The validity of this method also been testified using Lena gray image. Finally, according to the inconformity of recovered image via TVQC and TVEQ in pixel lever induced by image stitching, this paper presents a method to decrease the gray differences between recovered imag e blocks. Such method can effectively improve the PSNR value of recovered image, which provides a effective reference in high-resolution infrared CS imaging.