In order to research the on-board data processing for infrared (IR) scan images in Geostationary Earth Orbit (GEO) satellite, processing flows of data processors in the Phase Ⅱ and DSP-Ⅰ satellites for U.S. Defense Support Program (DSP) are discussed. A point target detection algorithm for IR scan images is proposed based on a two-channel filter to suit for space conditions. Firstly, the background prediction is modeled by a mean filter to suppress background clutters, and the adaptive threshold is determined on the residual image after the background elimination. Then, the peak detection is used to detect the peaks in IR images to reduce probability of false alarm originating from the adaptive threshold. Finally, the target identification algorithm using fusion technique is performed for the data from two channels. Experiments show that the proposed algorithm can get high detection probability and low false-alarm probability, and is easy and convenient for real-time operation. The obtained results indicate that the detection probability reaches 99.3% (the false alarm probability is 1.3×10-3) when the input Signal-to-Noise ratio (SNR) is no less than 6. The real-time analysis of the algorithm shows that the data processing capability can achieve 56.45 Mb/s. The proposed algorithm can meet the need of space-based data processing.