Shot change detection on news videos using color histogram and edge based approaches

Nowadays, video contents are stored in huge files or it is televised in continuous streams but the users need to retrieve only a particular part of topic or shots from video. So we need to split it in to short video segments with respect to appropriate retrieval units automatically and efficiently. In this research work, based on shot changes two different approaches are to detect shot boundaries. First approach is edge-based segmentation of shot change detection and second approach is color based histograms. These approaches are very suitable for detecting shot changes in large size news videos quickly in effective manner. The NDTV and TV9 news videos are used to detect shot changes with various time periods. The detected shots like anchor shot and non-anchor shot are stored in database. The experimental result shows that edge-based shot change detection performs well in detecting shot frames from given news videos.

I. Introduction

News videos are large in number. For segmenting the news in to single story units, an automatic and effective tool is required by the user. Shot boundary detection in video is also known as partitioning of video or segmentation of video which is the most fundamental step in retrieval of video and video data management indexing. To identify and classify different shot transitions in video sequences, the given video is segmented or split into essential shots. Various algorithms are suggested on the basis of edges [1], pixel color difference [2], motion [3]–[5], color histograms [6], [7], color ratio histograms [8] and predefined models, regions, objects or subsampling of spatio-temporal frame to detect camera breaks while taking videos [9]. This work illustrates the problem of video partition using color histogram and edge based approaches. Ordinary color problems, edges being mostly invariant under the local illumination changes, panning, tilting and also things affected by motion causes problem in detecting shots. In order to guarantee the robustness, both edge and color based shot detection is used in this framework. Time consuming problem is avoided with this multiple decompositions. Video indexing and segmenting are the problems relevant to shot classification and it is an important issue in video query browsing [10], [11], video compression, video traffic modeling [12] and video communication areas.