A video extraction method and device. The method includes acquiring original image resolution S0 of a Tth frame in which a target image corresponding to a target object is located, where T is a positive integer, and acquiring a resolution parameter P of the target image in the Tth frame, obtaining extraction resolution S1 according to the original image resolution S0 and the resolution parameter P, and extracting an image of the target object in a (T+1)th frame according to the extraction resolution S1. In this way, when extraction processing is performed on the target object in the (T+1)th frame, extraction processing is performed on the target image corresponding to the target object in the (T+1)th frame with reference to the extraction resolution, which is obtained through analysis, of the (T+1)th frame.