Abstract: In order to research the desired imaging spatial resolution of IR targets under different discrimination levels, including detection, recognition, and identification, a new method based on feature entropy, which is used to estimate IR imaging operating spatial resolution, was proposed. Firstly, IR target images under different resolution grades were obtained by down sampling. Secondly, IR target features extracted, including texture, shape and global statistical features, were normalized to the corresponding feature spaces to obtain the feature histograms. Thirdly, feature entropies of IR targets were calculated in order to measure the information of targets. Finally, according to the set thresholds, the desired lowest imaging spatial resolutions of IR targets under different discrimination levels were obtained, and taking the computer simulated IR targets for example, the presented method was demonstrated. The experiment results show that the conclusions are in good agreement with identification results, which can provide reference for optimization design and analysis of the infrared recognition system.