摘要
针对进化多目标图像分割算法运行时间长且依赖人工挑选最优解的不足,提出一种快速自动多目标图像分割算法。首先使用自适应Mean-shift算法对图像进行预处理,将粗分割结果进行二次分割以提高运行速度;其次选择相互排斥的指标作为多目标的目标函数,并采用RM-MEDA框架对超像素颜色与纹理特征分别进行优化,同时对它们使用不同权值作为目标函数优化;最后由模糊模型从众多Pareto折中解集中自动选择满足实际分割要求的PS解。引入Mean-shift进行预分割,相对于标准的RM-MEDA,其运行速度提高近18%,由模糊模型推荐的Pareto解中,97%的情况符合分割要求。
The multi-objective image segmentation algorithm is confronted with some challenges such as long running speed and manu⁃al selection of the optimal solution.In this paper,a rapid automatic multi-objective segmentation algorithm(RAMOSA)is proposed to solve this problem.Firstly,the adaptive mean-shift algorithm is used to pre-segment the image,and the coarse segmentation results are re-segmented to improve the segmentation efficiency.Secondly,mutually exclusive indexes are selected as the multi-objective function,and RM-MEDA framework is used to optimize the color and texture features of super pixels respectively.Finally,the fuzzy model automatically selects the specific optimal solution that conforms to the current situation from many Pareto results.Multiple image materials are selected for image segmentation experiment,the experimental results show that RAMOSA algorithm has higher segmenta⁃tion efficiency compared with general multi-target image segmentation.Mean-shift is introduced for pre-segmentation.Compared with the standard RM-MEDA,the operating speed is increased by 18%,and the accuracy of the results selected by the fuzzy model reached 70%.
作者
高华
邬春学
GAO Hua;WU Chun-xue(Office of Educational Administration,University of Shanghai for Science and Technology;School of Optical-Electrical and Computer Engineering,University of Shanghai for Science and Technology,Shanghai 200093,China)
出处
《软件导刊》
2020年第11期212-216,共5页
Software Guide
基金
上海市科学计划项目(16111107502,17511107203)。
