The line spread function (LSF) testing is the most important part of image quality assessment of optical imaging system. The measuring precision of LSF straightly affects the modulated transfer function (MTF) curve of system. A LSF testing algorithm based on geometrical projection is proposed to raise measuring precision. The LSF fitting curve is gained by analyzing the algorithm, cutting region of interest (ROI), identifying narrow slit and computing projection. The algorithm is validated by imaging narrow slit with background of a blackbody using a medium wave infrared detector. The results are compared with that using traditional way. The impact of un-uniformity noise on measuring results is analyzed. The main error of the algorithm is analyzed. It is concluded that the algorithm raises the measuring precision and the main factor of measuring error is the signal to noise ratio (SNR) of imaging system.