• 主题
高级搜索  >
历史搜索 清空历史
首页> 外文会议>2016 International Conference on Electrical Power and Energy Systems >AERASCIS: An efficient and robust approach for satellite color image segmentation
【24h】

AERASCIS: An efficient and robust approach for satellite color image segmentation

机译：AERASCIS：一种有效且强大的卫星彩色图像分割方法

获取原文
获取原文并翻译 | 示例
页面导航
  • 摘要
  • 著录项
  • 相似文献
  • 相关主题

摘要

Satellite color images carry a vast amount of information which needs an efficient image segmentation method to analyze. Because of its simplicity and low complexity, K-Means algorithm is frequently adopted for color image segmentation. But, usually the results of K-Means algorithm suffers from noises and hence over segmentation. This is due to the reasons that K-Means works on the basis of random K initialization and “Euclidean Distance Metric” as default. Also, in the case of satellite color image segmentation local contrast management is an important issue which is not paid attention in the traditional K-Means algorithm. So, in this paper, these problems are taken into consideration and a robust method has been proposed to tackle the same. First of all, HSV color space is chosen for color based transformation and calculations. Here, a Binary Search Based CLAHE is introduced for local contrast management. An entropy based technique is developed for determining the total number clusters and detection of initial centers of the clusters. “Cosine Distance Metric” is employed for distance based calculations involved in K-Means algorithm. The performance of the proposed approach is found robust with respect to noise and over-segmentation is removed up to a satisfactory level.
机译：卫星彩色图像包含大量信息，需要有效的图像分割方法进行分析。由于其简单性和低复杂度，K-Means算法经常被用于彩色图像分割。但是，通常，K-Means算法的结果会受到噪声的影响，因此会造成过度分割。这是由于K-Means基于随机K初始化和默认设置为“欧氏距离度量”的原因。同样，在卫星彩色图像分割的情况下，局部对比度管理是一个重要的问题，在传统的K-Means算法中并未引起注意。因此，本文考虑了这些问题，并提出了一种健壮的方法来解决这些问题。首先，选择HSV颜色空间用于基于颜色的转换和计算。这里，引入了基于二进制搜索的CLAHE用于局部对比度管理。开发了一种基于熵的技术，用于确定总数簇和检测簇的初始中心。 “余弦距离度量”用于K-Means算法中基于距离的计算。发现所提出的方法的性能相对于噪声是鲁棒的，并且过度分割被去除到令人满意的水平。

著录项

相似文献

  • 外文文献
  • 中文文献
  • 专利
获取原文

客服邮箱：kefu@zhangqiaokeyan.com

京公网安备：11010802029741号 ICP备案号：京ICP备15016152号-6 六维联合信息科技 (北京) 有限公司©版权所有

  • 客服微信

  • 服务号