Skip to main content

Characterization of polynomial filters for image enhancement and the extraction of invariant image features


Research objectives and content
The objective of the proposed research project is the inclusion of methods for image enhancement using polynomial filters in an already developed nonlinear adaptive system for invariant pattern recognition. The link between low and high level image processing consists in the similar structure of the polynomial filter kernels used for image enhancement and for the invariant feature extraction method. To achieve an effective combination the focus of the project will be on the derivation of a systematic characterization of polynomial filter kernels suitable for certain tasks by including a priori knowledge about the patterns in a given application. Together with the adaptivity of the pattern recognition system the proper choice of the filter kernels will lead to a substantial increase in robustness and separation ability, and a reduction of computational costs compared to general purpose image processing systems. Training content (objective, benefit and expected impact)
The design of the polynomial filters for image enhancement is based on Volterra theory whereas the filters used for the invariant feature extraction method are based on invariant theory. From a combination of these two approaches synergy effects are expected for the construction of a fast adaptive invariant image processing system as well as for the fields of image enhancement and invariant pattern recognitio individually.

Funding Scheme

RGI - Research grants (individual fellowships)


Università degli Studi di Trieste
Via A. Valerio 10
34100 Trieste

Participants (1)

Not available