Multimodal Content Based Video Retrieval System

Abstract

Content-based multimedia retrieval of video data has become
a important and challenging issue. Video mainly contains
several types of audio and visual information which are
difficult to extract, combine in common video retrieval
information. We present a video retrieval model that
integrates the utilization of various approaches. The proposed
system splits the videos into a sequence of shots and extracts a
very small number of representative frames from each shot
and it classifies them according to the various features. The
elementary video shots are extract by employing Clustering
for classification, low level feature extraction of colour by
HSV colour space conversion, shape by Canny Edge
Detection and texture by GLCM and audio processing by
Windowing method. The different features extracted using the
above approaches are stored in feature library. We propose
Euclidean distance, Clustering and spectrum matching with
windowing for similarity measure. In the proposed system, the
related videos are retrieved from the video database. On the
basis of different query clip, the videos are retrieving in our
system. In the learning and retrieval phase, the user provides
feedback on the relevance of each video searched among the
top returned searches.