In the matching process of multi-sensor images based on features, the incomplete contour caused by imaging may increase the difficulty in image matching. To solve this problem, a novel image matching method based on block shape features is proposed. Firstly, the contour features are extracted from a reference image and a real-time image respectively and are divided into blocks hierarchically. Then, the block features are extracted and normalized. Finally, the weighted similarity measure is used to implement matching localization. Because the area information of a candidate target is used to exclude false alarms, the correctness and robust of the matching are further improved. The method is tested in the matching of infrared and visible images. The result shows that the proposed method has better matching performance for multi-sensor images.