S.V. Basavaraja, S. Velusamy, and S. Varadharajan (India)
Video segmentation, Shot boundary detection, LZW.
We have proposed a efficient and robust shot segmenta tion algorithm that is based on a well known Lempel-Ziv Welch (LZW) text compression technique. The algorithm works on the fact that, when a frame is compressed using its previous/reference frame information, the ‘compression ratio’ obtained gives the direct measure of similarity be tween them. The novelty of the algorithm lies in, i) ex ploiting the characteristics of LZW technique in capturing the primary image features such as color and texture in formation, and representing them as a single feature called patterns; ii) easy handling of inter shot video effects such as abrupt, dissolves, fades, wipes, and intra shot video ef fects such as object entry/exit, slow/fast motion; and iii) efficient identification of shot boundaries and low compu tation complexity of the algorithm. The performance of the proposed algorithm is tested with a wide varieties of videos from a set of standard databases and with our lab video col lection. The experimentation results promises a high accu racy and robustness in identifying the shot boundaries.
Important Links:
Go Back
