Query by video clip

Typical digital video search is based on queries involving a single shot. W e generalize this problem by allowing queries that involve a video clip (sa y, a 10-s video segment). We propose two schemes: (i) retrieval based on ke y frames follows the traditional approach of identifying shots, computing k ey frames from a video, and then extracting image features around the key f rames. For each key frame in the query, a similarity value (using color, te xture, and motion) is obtained with respect to the key frames in the databa se video. Consecutive key frames in the database video that are highly simi lar to the query key frames are then used to generate the set of retrieved video clips. (ii) In retrieval using sub-sampled frames, we uniformly sub-s ample the query clip as well as the database video. Retrieval is based on m atching color and texture features of the subsampled frames. Initial experi ments on two video databases (basketball video with approximately 16,000 fr ames and a CNN news video with approximately 20,000 frames) show promising results. Additional experiments using segments from one basketball video as query and a different basketball video as the database show the effectiven ess of feature representation and matching schemes.