Evolutionary wavelet-based similarity search in image databases

There exist numerous image retrieval systems perform a fast similarity search in the image databases, but the quality of the outcomes provided by histogram-based image search is usually rather limited. One of the reasons is that histogram-based systems only support some form of global similarity using the whole color histogram as one vector. Meanwhile there is more information in a histogram than the distribution of samples resp. colors. In this paper, we first introduce some generic similarity search methods in the image database and then use them to form a generalized model that can adapt various transformations. Farther on, in virtue of the effectiveness of the evolutionary computation and based on the model, we construct a fitness function using the wavelet transformation to optimize the image similarity search by the particle swarm optimization. The results of our approach present a much better quality than other existing similarity search measures.