NICE-MSF Joint Seminar

Evolutionary Multi-Objective Optimization of Trace Transform for Invariant Feature Extraction

When?

Thursday 1 December 2011, 15:30 to 16:30

Where?

39BB02

Open to:

Public, Staff, Students

Speaker:

Mr Wissam Albukhanajer

The Trace transform is one representation of the image uses different functionals applied on the image function. When the functional is integral, it becomes identical to the well-known Radon transform, which is a useful tool in computed tomography medical imaging. One of the key questions in the Trace transform is the selection of the best combination of the Trace functionals to produce a triple feature as a robust image identifier. We employ an evolutionary algorithm (EA) inspired from nature and the biological reproduction to derive the best functionals in the Transform combined with the optimum number of projections. This achieved by multi-objective optimization (MOO) to calculate two cost functions, to minimize the within-class variance and maximize the between-class scatter. Pareto-based Non-Dominated Sorting Genetic Algorithm (NSGA) is adopted to find more than one optimum solution. The method is validated through experiments on various images from fish database. The proposed algorithm has given promising results with extremely efficient clustering. The algorithm is computationally inexpensive due to the efficient implementation and the very low feature dimensions compared to other methods exist in literatures.