An Approach for Resolving Occluded Data Based on Stochastic Optimization Techniques
Ganesh Vaidyanathan S, Kumaravel N and Bibhas Kar
This paper proposes a technique to identify data that occlude each other. In a secure communication system, wherein a data is sent in a partially hidden form, or in a problem like identification of palimpsests, we come across the challenge of finding partially hidden data. The proposed approach exploits techniques like evolutionary algorithm and particle swarm optimization. In both these approaches, a random population of solutions is extracted from the occluded data and by subjecting them to the genetic operators or the rules defined by the swarm intelligence, the partially hidden data is found out after certain number of generations. The number of generations required for identifying the hidden data and the probability of success depend on the amount of occlusion. The techniques work well even if up to 90% of the data is hidden. A comparison of performance both methods has been carried out and it is found that the proposed techniques give good results in terms of probability of success and accuracy of identification, though there is some compromise required on time required for arriving at the results.
Keywords: Occluded data, similarity index, rank of match, fitness value, swarm size, database, multi-point occlusion.