Deterministic Defuzzification Based on Spectral Projected Gradient Optimization

Authors:
Tibor Lukić;Nataša Sladoje;Joakim Lindblad
Affiliations:
Faculty of Engineering, University of Novi Sad, Serbia;Faculty of Engineering, University of Novi Sad, Serbia;Centre for Image Analysis, SLU, Uppsala, Sweden
Venue:
Proceedings of the 30th DAGM symposium on Pattern Recognition
Year:
2008

Citing 4
Cited 1

Algorithm 813: SPG—Software for Convex-Constrained Optimization

ACM Transactions on Mathematical Software (TOMS)
Nonmonotone Spectral Projected Gradient Methods on Convex Sets

SIAM Journal on Optimization
Feature based defuzzification in ℤ2 and ℤ3 using a scale space approach

DGCI'06 Proceedings of the 13th international conference on Discrete Geometry for Computer Imagery
Feature based defuzzification at increased spatial resolution

IWCIA'06 Proceedings of the 11th international conference on Combinatorial Image Analysis

Coverage segmentation based on linear unmixing and minimization of perimeter and boundary thickness

Pattern Recognition Letters

Quantified Score

Hi-index	0.00

Visualization

Abstract

We apply deterministic optimization based on Spectral Projected Gradient method in combination with concave regularization to solve the minimization problem imposed by defuzzification by feature distance minimization. We compare the performance of the proposed algorithm with the methods previously recommended for the same task, (non-deterministic) simulated annealing and (deterministic) DC based algorithm. The evaluation, including numerical tests performed on synthetic and real images, shows advantages of the new method in terms of speed and flexibility regarding inclusion of additional features in defuzzification. Its relatively low memory requirements allow the application of the suggested method for defuzzification of 3D objects.