Performance Improvement of Content-Based Video Retrieval using Thepade’s Sorted n-ary Block Truncation Coding(TSnBTC) with Various Color Spaces

Abstract

Content-Based Video Retrieval is used in several applications across sectors such as Education, Medical, and Agriculture. In this world, each end-user has a different perspective when elaboration of a video is expected, due to which the text description/keywords-based methods of video retrieval do face difficulties. Content-Based Video Retrieval (CBVR) has become more popular in the last few years; which attempts to describe each video, based on the contents represented as features. Various feature extraction methods can be used for CBVR. This paper attempts CBVR with the features extracted from videos using TSn-BTC - Thepade’s Sorted n-ary Block Truncation Coding n-ary method along-with different color spaces. The different types of an extended version of TSnBTC are enlisted in this paper ,Versions are - Thepade’s Sorted Ternary, Quarternary, Pentanary, Septanary, Octanary, Nonary, and Denary Block Truncation Coding Technique. Plenty of color spaces used in this paper such as RGB, LUV, YCbCr, YIQ, YUV, CIEXYZ, CIEXYZrec, CIEXYZccir, CMY, and CMYK, in Which YCbCr gives Higher Accuracy followed by RGB. The similarity criteria used is Mean Squared Error(MSE) for calculating similarity. The Feature Vector generation process can be done by using the 20 <sup xmlns:mml="http://www.w3.org/1998/Math/MathML" xmlns:xlink="http://www.w3.org/1999/xlink">th</sup> Frequency Frame of every video. The experimentation environment in Content-Based Video Retrieval based on TSnBTC Technique is done of 500 various videos which consist into Various 10 unique categories. Each category consists of 50 videos. The YCbCr Color Space achieved higher performance. For each query video fired onto no .of database Video dataset, and the Mean Exactness is calculated For different TSnBTC Versions.