Image processing includes several steps that segmentation is the most important step of the procedure. Segmentation is the phase in which inputs get separated into their components that assign long time. One of the most basic methods of segmentation is presented by graph theory. According to the theory, each node in a graph is a representative of a pixel in the picture and each edge joint's adjacent pixels. Weight corresponding to each edge is based on some properties of primary and terminal pixels of the edge. On the other hand, graph partitioning refers to graph nodes categorization to two or more parts based on certain criteria. Up to now image segmentation is performed by optimized techniques such as genetic algorithm, ant colony, … statistics and graph-based methods. In this article, to resolve the issue of image segmentation, input image converts to graph after initial pre-processing. It obtained graph is partitioned with imperialist competitive algorithm, and the amount of crossing edges optimizes through the graph. Afterwards, this graph applied to the image. Therefore, it divides into sections. Berkeley Segmentation Dataset images have been utilized in order to survey the resulting solution. Statistical results indicated that in approximately 90% of cases. The imperialist competitive algorithm has achieved better results.
展开▼
