An Image Retrieval Algorithm Based on Improved Electromagnetism-like Mechanism

摘要

In view of the advantages, such as the simplicity for understanding and high convergence rate of Electromagnetism-like Mechanism (EM) algorithm, an image retrieval algorithm based on improved EM is proposed in this paper. The novel algorithm assigns distinct weights to each block of the image, which represents the extent of human eyes' preference to different areas when recognizing this image. The Chaos Search (CS) is used in local search in order to prevent the retrieval process from falling into a local optimum leading to a premature convergence. A genetic coefficient is added to the updating formula so that the particles are more likely to move into other feasible regions. The experimental results demonstrate that the improved EM algorithm can figure out the best matching image rapidly and accurately.