Research On Bee Colony Algorithm And Its Applications In Image Processing

Bees Algorithm (BA) is an optimization algorithm based on swarm intelligence developed by Karaboga in 2005. BA is a kind of bio-inspired computing, and designed by simulating the actions that honey bee colony explore for high quality of food sources. The excellent feature of BA is that the local searching and global local searching work at same time in each iteration, and it also has some other merit such as fast computing, perform easily, and so on. BA has become a new research hot point in computation intelligence area for its rapid convergence and simple to implementation, while the theory research is still in initial step stage. In the thesis, deep theory analysis for BA is made, and improved algorithm is proposed and applied it into digital image processing. Several points are included in this paper as follows:1 The analysis is performed on BA. A strict mathematical description is given for the basic conception of BA algorithm, with the definition of notion, such as bee state, bee colony state, and so on. The bee state transition equation has proved to be a Markov chain process, based on which, an elementary convergence analysis of BA has finished. The BA is introduced in the optimal problem of multiple peak and high dimensional function, and a lot of experiments have been done for testing that BA has ability of convergence.2 According to the analysis result of BA, the improved method of BA is proposed. First, because the leader bee can remember the historical optimal place, the directional information, which between two historical optimal places that they are on different time, should be merged into the searching strategy. In other word, the leader bee can select the next explore point accord to the directional information. Second, the radius of searching area is modified that radius can decrease gradually in the iteration, but when the leader bee trips out the local optima, the searching radius can adaptively become big at the same time. The improved BA can adapt to searching in the large scare area and fine area.3 BA is introduced into image segmentation. The image segmentation by single threshold based on BA and fuzzy entropy is proposed. The fuzzy membership function is set for image pixel classified, the searching variance in fuzzy space is simple from three to one, so the amount of computing is decreased and the velocity of computing is faster. The image segmentation by multi-threshold based on BA and PSNR is also proposed. PSNR is applied to segment image firstly in this method.4 BA is also introduced into data clustering. In multi-dimensions data space, there may be large difference in value‘s range among different dimension. Uniform searching radius can not satisfy searching task. For this problem, a flexible radius selected strategy is proposed, and it is applied successfully on image data clustering.5 The BA is focused on researching on removing noise in image. A method of removing pulse noise in image is proposed. The method has two steps: the fist step, the image pixel contaminated by pulses noise is checked out; the second step, the contaminated pixel is restored by using BA. Experiment results show that the method has powerful ability to remove pulse noise.6 The BA is also focused on researching on extracting image edge. A lot of local optima arise in the searching global optima process of BA, these local information can be used to extract some points of edge. The other edge point can be searched from these points by new searching rule.