To meet the objection identification need under a low-speed data chain, a coding strategy about a Region of Interest (ROI) is introduced into infrared video coding. On the basis of a H.264 video encoding/decoding framework, the ROI quantitative model is optimized by adding the processing of ROI encoding, constructing the video encoding/decoding framework and adjusting the quantitative parameters of ROI and non-ROI macroblocks in the bit rate control process. The experimental result shows that this method can save the limited bit rate, increase the details of ROI objects, improve the definition of ROI objects and improve the subjective visual quality on the whole.