A novel approach to the problem of visible and infrared automatic image registration was proposed. The registration is performed by extracting affine covariant regions through same level extremal region (SLER) detector on a gray gradient image. Then, hypergraph matching algorithm was employed to obtain identical key points. The approach is especially suitable for registering multi-sensor infrared images where the quality of images or the corresponding edge maps are worse than the counterparts on a common optical image. Experiments performed on several challenging real image pair show that our proposed method achieves better performance than other approaches.