Skip to Main Content
In this paper, an approach to optimization of semantic image analysis is presented by employing genetic algorithm coupled with ontologies. Ontologies are used to interpret the image in machine understandable language. High-level information is represented in the form of interested domain concepts chosen and the low level information in the form of low level visual descriptors. These low level descriptors are extracted from the segmented image and visual similarity is assessed in terms of degree of confidence, which forms initial hypothesis set. This is further passed into genetic algorithm along with extracted spatial relations for most optimized annotation. Experiments with a collection of images belonging to a specific domain demonstrate the performance of the proposed approach.