Objetivo
This project consists of development and analysis of high-quality and time-efficient algorithms for image data compression by using hierarchical segmentation. The hierarchical segmentation is based on multi-stage approximation of image homogeneous regions by low-degree polynomials as well as on detection and coding of edge segments as bilevel image segments. They have to be compactly encoded by contour piecewise-linear approximation and/or vectorisation procedures. It is proposed to use adaptive post-filtering algorithm, steerable by error of polynomial approximation, in order to diminish the error of the reconstruction of edge segments and to smooth the block effect at the decompression phase.
For fast implementation of image adaptive filtering, edge detection and region approximation, a unified approach is proposed. This consists of a preliminary approximation (or exact representation) of the local processing function (operator) in the best way by simpler operators, which allow recursive computations with both temporal and spatial recursion. It is proposed to use separable two-dimensional processing of the image with spatial recursion in two orthogonal directions. Parallel implementation allows for real-time implementation of the developed techniques and includes realisation in workstation clusters.
The main application area will be fast segmentation and compression / decompression of images from medical imaging.
Tema(s)
Convocatoria de propuestas
Data not availableRégimen de financiación
Data not availableCoordinador
5020 Salzburg
Austria