Content Based Video Retrieval using Local Ternary Pattern Feature

Abstract

In the 21st century, image and video repositories have been increasing drastically.This is attracting visual based smart city solutions for transport, healthcare, healthcare,safety, hospitality, sports visuals, etc. Effective storage of Video database andits retrieval is the new problem to solve. This search analyzes metadata like keywords,captions, titles, etc. With current multimedia solutions of computer visionand advanced image processing, the CBVR approach uses content understanding ofimages like color, shape descriptor, textures, deep features, etc. Limitations in inheritanceof metadata systems have led to content understanding based approaches.This work proposes to generate feature vectors for each of the data base and queryvideos. We�ve proposed to use a color feature based PCC distance for video shot detectionand key frame extraction to remove redundancy or dimensionality reduction.Further uses local ternary pattern (LTP) and uniform local ternary pattern dynamictexture feature on key frames and feature vector generation. Then Euclidean distancewith KNN classifier for video retrieval. We�ve utilized UCF50 human actiondata set for the proposed work.