Research On Application Of Improved Firefly Algorithm In Image Enhancement And Segmentation

Due to the complexity and huge mount of data of images, in general, the traditional techniques for image processing and analysis have very high computational complexity. As a result, the image processing techniques based on evolutionary computation and have been attracted much attention. Firefly algorithm(FA) is a newly proposed meta-heuristics algorithm, which needs few parameters and has strong global optimization ability. It has been applied in some problems in the field of image processing; however its computational efficiency is not high. Therefore, the clustering firefly algorithm based on levy flight(CBLFA) is proposed and utilized for image enhancement and segmentation in the thesis, main work is as following.Firstly, CBLFA is tested on some well known benchmark function optimization problem.The experimental results show that CBLFA algorithm has stronger global optimization ability and faster than the basic FA.Secondly, the adaptive image enhancement method based on CBLFA is studies. Considering that the incomplete normalized Beta function is able to fit curve of typical gray-scale transformation and the optimal fitted curve parameter- alpha and beta might be treated as combinatorial optimization problems, CBLFA is used to solve the problems to achieve image enhancement adaptively. Experiments show that the method can be applied to all kinds of degraded images; the enhanced images have more uniformly distributed gray scale, better gray scale contrast and more clear details.Finally, thresholding approaches based on 1-D histogram and 2-D or 3-D histogram optimized with CBLFA is explored and implemented. For instance, multiple thresholding methods based on Otsu are solved with CBLFA. Furthre, 2-D Ostu and 3-D Ostu thresholding methods optimized with CBLFA are also fulfilled in the thesis. In addition, the basic FA and particle swarm optimization(PSO) algorithm are also conducted on above applications. Results show that CBLFA can quickly obtain the suitable thresholds, significantly reduce the execution time and could obtain the better fitness function value than basic FA and particle swarm algorithm on the whole.In sum, the improved FA and its application for image enhancement and segmentation is studied in the thesis, the adaptive image enhancement algorithm and multiple thresholding methods based on CBLFA are carried out. The contrast experiment of application performance is conducted with basic FA and basic particle swarm optimization. The experimental results displays that the proposed algorithm proforms well and has some potential applications in the field of image processing.