4 May 2012 Query by example video based on fuzzy c-means initialized by fixed clustering center
Author Affiliations +
Currently, the high complexity of video contents has posed the following major challenges for fast retrieval: (1) efficient similarity measurements, and (2) efficient indexing on the compact representations. A video-retrieval strategy based on fuzzy c-means (FCM) is presented for querying by example. Initially, the query video is segmented and represented by a set of shots, each shot can be represented by a key frame, and then we used video processing techniques to find visual cues to represent the key frame. Next, because the FCM algorithm is sensitive to the initializations, here we initialized the cluster center by the shots of query video so that users could achieve appropriate convergence. After an FCM cluster was initialized by the query video, each shot of query video was considered a benchmark point in the aforesaid cluster, and each shot in the database possessed a class label. The similarity between the shots in the database with the same class label and benchmark point can be transformed into the distance between them. Finally, the similarity between the query video and the video in database was transformed into the number of similar shots. Our experimental results demonstrated the performance of this proposed approach.
© 2012 Society of Photo-Optical Instrumentation Engineers (SPIE)
Sujuan Hou, Sujuan Hou, Shangbo Zhou, Shangbo Zhou, Muhammad Abubakar Siddique, Muhammad Abubakar Siddique, } "Query by example video based on fuzzy c-means initialized by fixed clustering center," Optical Engineering 51(4), 047405 (4 May 2012). https://doi.org/10.1117/1.OE.51.4.047405 . Submission:

Back to Top