Current Trends in Video Search

We’ve seen that video is ubiquitous on the Web, and that video search sites may focus on a particular aspect such as video sharing, aggregation or original content. Regarding search, sites may have a rich tagging capability or support queries similar to traditional databases while others support full text search of dialogs. Our examination of a wide array of video sources intended for various applications brought to light the range of available metadata that may be available for a given class of video material. Some are rich in standardized metadata, while other content sources such as user-contributed have limited metadata and may rely more on feedback from viewers to enrich the metadata and provide ratings so that others may navigate the content more easily. We’ve reviewed basic digital video technology with a focus on issues related to video search such as compression, transport, metadata representation, container formats, and media player systems. Our introduction to media processing presented the state of the art in metadata extraction from media with the goal of augmenting available metadata for retrieval applications. We examined processing individual media types including text, audio, and video as well as multimodal techniques. Theses methods are embodied in the research systems which have been improving steadily over the years.

In this chapter, we observe certain trends related to video search. We will try to avoid predicting the future, but by articulating where things were and the current state of the art, we can give the reader a sense of the general direction of constituent technologies as well as social trends. Increasing bandwidth, storage density and computational power are familiar to us given Moore’s law and they enable advances such as the cost effective evolution from standard definition to high definition video – even for consumer camcorders.