Summary of the research project

The project “New perspectives on intelligent multimedia management with applications in medicine and privacy protecting systems” is intended to develop fast methods for processing and search of multimedia data in the databases of a very large number of records and implementation of these methods in relational database environment. The proposed set of solutions will enhance security for storage of confidential and sensitive (e.g. biometrics) data, and communications related to it. Realization of the objectives of the project, extremely important from the point of view of the EU, including the software implementation, requires addressing important and challenging research problems. The heart of the project is to solve these problems and apply groundbreaking methodology for efficient retrieval images on the basis of their content (CBIR) and embedding them in a relational database management system.

The main challenges

challenges
The proposal consists in six challenges:

  • Challenge 1: Development of invariant and robust image descriptors (features, keypoints)
  • Challenge 2: Image representation (called fingerprint) generation, based on extracted image features
  • Challenge 3: Design of a new type of field in relational databases for efficient fingerprint storage
  • Challenge 4: Design of new type of index in relational databases for efficient multimedia retrieval
  • Challenge 5: Development of identification, retrieval and classification techniques based on new database features
  • Challenge 6: Design of a new standard of SQL language for multimedia retrieval based on full-text search technology
To prove the usefulness of the methods developed throughout the project, two real world applications will be developed – renal medical imaging retrieval and privacy preserving search in large scale distributed systems. The project will be carried out jointly by the Department of Computer Engineering at the Czestochowa University of Technology and Stochastic Information Processing Group at the University of Geneva in the span of two years (2011-2012). The expected overall success is based on the fact that the methods that will be developed in the project will rely on deep synergy of project partner expertise in image processing algorithms (Swiss partner) and computational intelligence techniques (Polish partner).

The main milestones

The main milestones of the project will be:

  • a. Novel technology for fast search in multimedia databases
  • b. New method for fingerprint generation
  • c. Design and implementation of a new type of field for fingerprint storing
  • d. Dynamic library with code for operators (e. g. compare function) based on artificial Intelligence algorithms
  • e. Design and implementation structure of the SQL index type based on the R-tree for fields with fingerprints
  • f. Special SQL language extension to work with images and their fingerprints
  • g. Two applications
Efficient implementation of the ideas and research results conveyed by the project will provide a positive impact in both social and economic areas of the society. The ability of a fast search in multimedia databases based on a new type of image representation (called in the literature “fingerprint” or “robust hash”), without revealing the entire original content, will be a breakthrough in regard to respect and protect privacy sensitive information. This aspect is definitely of interest for authentication based on biometric data, where the need for techniques of authenticity verification, without the necessity to exploit privacy sensitive information, is highly demanded. Moreover, such a privacy-safe solution will increase the overall public acceptance for these methods. A similar social impact is expected in the medical domain with regard to patient-relevant data disclosure. The results of the project will be communicated by presenting papers in the most renowned journals and conferences, by a web page dedicated to the project and by seminars dedicated to the project to enable for the widest possible dissemination.