The invention discloses a video identification based system for extracting film images. The system is characterized by comprising the following steps of: firstly, converting video frequencies and audio frequencies by a pretreatment server into standard voices to be identified; secondly, converting the video frequencies by the pretreatment server into standard images to be identified; thirdly, converting video subtitles by the pretreatment server into standard binaryzation images to be identified; fourthly, carrying out feature extraction on the voices, the pictures in a picture sample database and the binaryzation images to be identified; and fifthly, carrying out feature extraction on the voices, images or characters of an identified video file by an index server, comparing the similarity of the obtained features of the voices, the images or the characters with that of feature values in the feature database which is obtained in the fourth step, and finally obtaining match results to be called by users. The system is characterized by accurately capturing the images, completely preventing the functions of identifying and extracting the film images from influencing the original webpage browsing speed and the like.