Ant Colony Algorithm And Its Application In Image Processing

Ant colony algorithm was proposed by Italy scholar Marco Dorigo etc. in 1990s,which is inspired by the behavior of ants when they search food. It is another new colony intelligent algorithm. Ant colony algorithm has some characteristics such as distributed computing, self-organization, positive feedback and so on. But it also has some shortcomings such as dis-convergence, finding the local solution and so on. The main content of research is as follows:Firstly, deeply researching the model of ant colony algorithm, summarizing the research progress about ACO in recent years and researching the application circumstance in image segmentation.Secondly, According to the characteristics of images, deeply analyzing the disadvantages of ant colony algorithm, then improving the model and proposing a model called Region-Edge ant colony algorithm. There are two kinds of ants in the improved model, one of which called region ants when another called edge ants. Different ants have different path choosing strategies and trial updating strategies. so the ants avoid to get into stagnation and can search the best solutions. The results of experiments indicate that the improved model has the advantage in computing efficiency and segmentation precision.Thirdly, According to the characteristics of intelligent searching and the whole optimization which is from ant colony algorithm and the advantage of being fit for processing real time images which is form the model of pulse coupled neural network(PCNN), proposing an improved model called ant colony-pulse coupled neural network(AC-PCNN). Traditional model of PCNN has a lot of parameters which are need to setup each time, segmenting different images needs setup different values to each parameter and different value of each parameter has great influence to the result of segmentation. Therefore, using ant colony algorithm to search the optimal value of each parameters in their solution space, the advance directions of ants are controlled by dynamic path choosing strategy and two kinds of trial updating strategies. The purpose of improved model is actualizing automatic PCNN system, which can segment different images automatically. Experiment results indicate that AC-PCNN algorithm is viable and the effect of finally segmentation is better.