Feature Matching would be effective here I think.
In the module feature2d you can find several Detectors/Descriptors which extract Features of a given reference Image.
After that you can Match them with Flanns Matching algorithms.
If you don't like the Detectors/Descriptors in features2d there are also additional Detectors/Descriptors in opencv-contribs xfeatures2d module. But be aware the Detector/Descriptors from feature2dx (e.g. SIFT and SURF) are still under a U.S. License and not under the normal opencv License.
