In recent years, the multimedia data have taken a leading role in various application fields such as the medical diagnostics, the protection and promotion of cultural heritage, the video surveillance, the fight against counterfeiting, the e-learning platforms.
The continuous generation of multimedia data requires effective tools for the exploration and analysis of visual information by the construction of relevant meanings between the features of the multimedia content.
Indexing and metadating (Textual or content-based) are essential to facilitate user access to the multimedia collections through efficient search engines for the creation of meaningful presentations that answer the user queries.
During 2015, FUB activities related to this context, have dealt with the:
Development of new automatic techniques for low level feature detection and representation following bag of words approach
Semantic labelling and classification of multimedia content using Machine learning methods
Development of prototype system, based on Hadoop and Spark technologies, for the management and processing of complex queries on the basis of the semantic descriptors.
FUB interest towards these issues is motivated by the opportunity to apply the developed methods to the protection of industrial ownership and copyrights.
With respect to audio signal, FUB has developed efficient methodologies to automatically classify the environmental sound, so that it has ensured the coordination of the European project “Octave” (Objective Control of Talker VErification).
Multimedia data management also includes visual information encoding and transmission to the end user that can result in quality loss of the multimedia signal. This has led FUB to continue its decennial activity in ITU and MPEG, about objective and subjective metrics for video quality assessment.