Robust and Efficient Transform Domain Video Sequence Analysis: An Approach from the Generalized Color Ratio Model
✍ Scribed by Donald A. Adjeroh; M.C. Lee
- Book ID
- 102976023
- Publisher
- Elsevier Science
- Year
- 1997
- Tongue
- English
- Weight
- 686 KB
- Volume
- 8
- Category
- Article
- ISSN
- 1047-3203
No coin nor oath required. For personal study only.
✦ Synopsis
selection of representative frames which form the basic Indexing is an important aspect of video database manage-index for each video partition. Further operations, such as ment. Video indexing involves the analysis of video sequences, video browsing, and content-based search and retrieval of which is a computationally intensive process. However, effecobjects in the video can then be performend using only tive management of digital video requires highly efficient, yet the representative frames, instead of the entire video. With robust indexing techniques. Based on the properties of the genthe huge volumes of data involved (even in compressed eralized color ratio model, the paper presents techniques for form), video sequence analysis 1 becomes very computarobust and efficient video indexing, by analyzing the video tionally intensive, and highly efficient but robust techsequence directly in the compressed domain. Illumination and niques are essential. Proposals have thus been made to motion invariant color ratio features, appropriate for blockperform the analysis directly on compressed video [5, 40, based transform domain video indexing are presented. Robustness is achieved by use of (i) illumination and motion invari-42], particularly, for those based on the Discrete Cosine ant color ratios; (ii) formally derived thresholds with predictable Transform (DCT). performance; and (iii) both DC and AC coefficients. Gains in In general, for both compressed and uncompressed efficiency accrue from: (i) operating only on a formally selected video, the basic mechanism used to reduce the amount of minimal set of blocks; (ii) performing all operations directly computation involves choosing only a subpart of the entire in the compressed domain; and (iii) avoiding computationally frame for processing. So far, the choice of what proportion expensive motion compensation procedures. Some experimenof a frame to be processed has been made randomly [5, tal results are presented which show the superiority of the 22, 39] without any formal basis. A random choice could proposed methods over traditional approaches. © 1997 Academic either underestimate or overestimate the appropriate Press amount of information needed to effectively discriminate between video scenes. Moreover, an arbitrary (or even a tion of the various editing effects in the video and the 182