改进核模糊C均值聚类算法在服装图像分割中的研究

Clothing Image Segmentation Based on Improved Kernel Fuzzy C-means Clustering Algorithm

图像分割是基于内容的服装图像检索系统的重要研究内容之一,传统的模糊C均值聚类算法在用于图像分割时,存在对样本分布和输入参数有限制、收敛于局部最优解、对噪声敏感等缺点.本文加入高斯核函数优化空间样本数据计算,并利用局部空间信息提高图像分辨能力,有效改善了模糊聚类算法.通过实验发现,改进的算法提高了图像类与类之间的分离性,分割结果更准确,对样本间的微小差别处理较好,具有更加可靠的性能.

Image segmentation is one of the important research contents of content-based clothing image retrieval system.The traditional fuzzy c-means clustering algorithm has some shortcomings on limiting,sample distribution and input parameters,converging to the local optimal solution and being sensitive to noise when it is used for image segmentation.In this paper,the Gaussian kernel function is used to optimize the spatial sample data calculation,and the use of local spatial information is to improve image resolution,effectively improving the fuzzy clustering algorithm.The experimental results show that the improved algorithm improves the separation between image classes and classes,resulting in more accurate segmentation,better handling the difference between samples and having more reliable performance.