Navigating News Programs in Large TV-Streams: A Knowledge-Based Approach

2. Related Works

Intensive research efforts were directed to content-based video indexing (Hu et al. 2011; Choroś, 2015b). The video documents are habitually incorporated in large TV streams. Therefore, efficient indexing of TV programs requires firstly their localization and extraction them from the stream. A class of works focuses on locating programs in TV stream, where as other kind of works is interested on the internal structuring of TV programs. Few efficient contributions have been proposed to structure TV streams. For example, to identify the various programs in TV stream, Naturel and Gros (2008) created hash-table containing key-frame signatures of program references. Then, he used signature similarity between the stream signal and the hash-table elements to locate the programs. The method presented by Manson (2009) aims to detect automatically programs in TV streams. The purpose of its work is to detect the limits of broadcasted programs. The method first detects inter-programs as repeated sequences in the TV stream in order to deduce the programs boundaries. The major drawback of these contributions is that they focus only on programs localization in TV streams without reaching the second level: the internal structuring where we can extract the semantics.