Querying Similar Images with Linear Superpositions [QI2012]
We have analyzed new technique for Information Retrieval (IR) making use of the geometric structure of Hilbert spaces in analogy with the Postulates of Quantum Mechanics (QM), in order to exploit the Physics approaches in multimedia contexts. In particular, we propose a method for the creation of a Vector Space Model (VSM) for representing digital images through the use of the MPEG-7 Visual Descriptors, enabling the evaluation of similarity scores, based on the analysis of the Postulates of QM. The proposed software tool creates clusters made up of superposition of image vectors that are taking into account the distribution of the feature values of all members and could benefit the current techniques for content based query applied to an information system.
Use of HSV for similarity search on digital images.
We are investigating MPEG-7 Visual Descriptors in a perceptive space making use of the formalisms of Quantum Mechanics. Some degree thesis (Pellegrino, Barresi, Battista) in collaboration with the Experimental Physics department of University of Torino are currently analysing this novel approach to the IR.
We have investigated and started new methodology for managing metadata on distributed hash table [Informatica2009], pioneering this approach that is currently widely used today.
- [QI2012] Elisa Maria Todarello, Walter Allasia and Mario Stroppiana - MPEG-7 Features in Hilbert Spaces: Querying Similar Images with Linear Superpositions - Quantum Interaction 2011 - Aberdeen June 2011. Published onto Sringer LNCS.
- [Informatica2009] Walter Allasia and Francesco Gallo and Marco Milanesio and Rossano Schifanella - Indexing and Retrieval of Multimedia Metadata on a Secure DHT - International Journal of Informatica: Special Issue on Multimedia Information System Security - Slovenian Society Informatika - Issue 33(2009), pages 85-100.