The invention relates to a method for video registration of a candidate video among a plurality of master video, comprising steps of extracting a fingerprint for candidate video and a fingerprint for each of a plurality of master video; identifying a master video among the plurality of master video which fingerprint best matches fingerprint for the candidate video; and registering the candidate video with the identified master video by estimating a distortion model between fingerprint for the candidate video and fingerprint for the identified master video.