Performance rise in Content Based Video retrieval using multi-level Thepade's sorted ternary Block Truncation Coding with intermediate block videos and even-odd videos

With the need of efficient video retrieval system, the content based approach is most important part of video retrieval system. As text based video retrieval is degrading its performance with respect to major issue of user's probabilistic perception to a video, so there is need of revision for video retrieval technique by content based style. Content Based Video retrieval (CBVR) is emerging system in any video retrieval applications. The Block Truncation Coding (BTC) [17, 18] is one of the color feature extraction methods in CBVR. The extended version of BTC is Thepade's sorted Ternary BTC (TSTBTC). This TSTBTC can be further stretched as multi-level TSTBTC and on even odd videos and also applying on intermediate blocks of videos as feature extraction method is proposed in this paper. The video data set considered is of 500 videos with 10 categories for experimentation. For testing, including RGB color space, other 6 color spaces are considered (KLUV, YIQ, YUV, YCgCb, YCbCr and XYZ). The similarity between query video and video from database is done by absolute difference (AD) measurement. The performance of method is confirmed by use of precision and recall. The average precisions are calculated for each considered video from database as query. Thus, for each color space average precision is calculated. The height of average precision and recall cross over point is considered hence, better is the height more accurate method is the feature extraction technique. In case of multi-level and even odd videos using TSTBTC, KLUV color space is performed well followed by YIQ color space. In case of intermediate blocks, the YIQ color outclassed followed by KLUV color space.