Performance gain of content based video retrieval technique using intermediate block truncation coding on different color spaces

The elementary technology for video retrieval technique in computing is very sensitive to the user´s query not to the contents of video. Every user has perspective to the video in the world of videos. So to identify this is difficult using text based video retrieval as this retrieval is based on text as processing input for video retrieval. The better approach for this video retrieval is content based video retrieval which is gaining acceptance in world of videos as a technique. In this paper, a novel approach for increasing the performance of retrieval by using intermediate levels with Block Truncation Coding (BTC) is presented. For this, the experimental data set is built up of 10 different categories in which 50 videos are present per category. In this paper, the approach of generation of feature by using 20^th frequency frames of videos from data set considered. Each of the extracted frame is intermediated by intermediate blocks (which is considered as 4 and 9) and then BTC applied on each block to get feature vector suing on 7 different color spaces as KLUV YCbCr, YCgCb, YIQ, YUV RGB, and XYZ. The feature vectors generated are processed and best match results are achieved. The similarity criteria used is absolute difference (AD) for calculating similarity or difference among videos and query. The performance of this technique is confirmed with help of cross over points of precision and recall values for each color space. The higher performance is achieved is in YIQ color space using intermediate block level 16.