An apparatus (MDS) and related method for image processing. The apparatus includes a multi-modal matcher (M). The matcher (M) operates to match a first reference (RF1) image against a stream (LF) of image frames (F1, F2). Once a match is found, the current reference image (RF1) is exchanged for a matching frame from the stream. The matcher (M) then attempts matching against subsequent frames in the stream by now using said matched frame as a new reference image. Once a further frame (Fm) is found that matches said new reference image, said further frame (Fm) is output as a best match.