Text this: Feature extraction and localisation using scale-invariant feature transform on 2.5D image