Texture Content Based Video Indexing and Retrieval using English Captions

Abstract

The project Content Based Video Retrieval (CBVR) developed to index the video that have English captions, it contain different modules for indexing the video that are text detection,text extraction and text classification, text detection system that is based on a combination of unsupervised and supervised techniques. The unsupervised technique is based on the image analysis techniques which determines the contrast, alignment of the text followed by geometrical properties of text and identify the text area in an image. Text areas are then validated by Support Vector Machine (SVM) using a combinations of features computed from Gray Level Co-occurrence Matrices (GLCM). After validation potential text area feed to the Optical Character Recognition (OCR). According to text returned by OCR frames are indexed to the database. Once the video is indexed user can query the database corresponding to that query frame will pop-up which matches the query, from that frames user can start playing the video.