Most pollutant gases have obvious absorbing or emitting features in the infrared band. Passive Fourier transform infrared (FTIR) remote sensing technology is used to detect and identify pollutant gases in a standoff distance. In the case of remote sensing pollutant gases on the mobile platform, the background is unknown and the spectrum includes interferents′ spectral features such as atmospheric gases, thus background compression method is needed to extract target spectral features. An infrared background compression method based on brightness temperature spectrum is proposed, in which the background radiance is referred as a slowly-varied baseline. The radiance transmitting simulation software named MODTRAN is used to simulate interferents′ spectral features such as atmospheric gases, and the infrared background compression is realized based on least-square fitting principle. The experiment takes ammonia as the target gas and the low altitude sky as the background. The results indicate that this method can effectively compress the background and interferents′ features, extract the ammonia′s spectral features and calculate the concentration-path-length.