Author Login Editor-in-Chief Peer Review Editor Work Office Work

Computer Engineering

Previous Articles     Next Articles

Image Classification Algorithm with p-voltages Based on Mean Shift

RUAN Yingying,WANG Xili,LIN Hongshuai   

  1. (School of Computer Science,Shaanxi Normal University,Xi’an 710119,China)
  • Received:2015-04-29 Online:2016-06-15 Published:2016-06-15

基于均值漂移的p电压图像分类算法

阮莹莹,汪西莉,蔺洪帅   

  1. (陕西师范大学 计算机科学学院,西安 710119)
  • 作者简介:阮莹莹(1987-),女,硕士研究生,主研方向为图像处理、模式识别;汪西莉(通讯作者),教授、博士、博士生导师;蔺洪帅,硕士研究生。
  • 基金资助:
    国家自然科学基金资助项目(41171338,61401265)。

Abstract: p-voltages algorithm is a semi-supervised classification algorithm,which is based on samples’ theoretical voltages in the electricity network system.It has some disadvantages in image classification,such as only two labeled samples (source node and sink node) are selected,resulting in poor accuracy of classification,and the high complexity of the design of graph makes it unable to process large images.To solve the above problems,image classification with p-voltages algorithm based on mean shift is proposed.Mean shift method is used to smooth the image for reducing the diversity of image features,and multi-labeled samples are chosen from the smoothed image as the source and sink nodes to improve the effectiveness of learning.A sample in each smoothed area is randomly selected as the unlabeled sample to ensure that it carries abundant image feature information.In order to reduce the scale of design of graph and furthermore provide conditions for large-scale image classification,the labeled and unlabeled samples are used as the subset of the original image for composition.Experimental results indicate that the proposed method not only improves the classification accuracy,but also receives a higher time efficiency,suitable for large-scale image classification with complex features.

Key words: p-voltages, electricity network system, semi-supervised, image classification, mean shift

摘要: p电压算法是电网络系统中基于样本理论电压的一种半监督分类算法,在图像分类中仅选取两个标记样本(源点和汇点),分类正确率较低,构图复杂度大,且不能处理规模较大的图像。为此,提出基于均值漂移(mean shift)的p电压图像分类算法。通过mean shift算法平滑图像以降低图像特征的多样性。在平滑图上选取多个标记样本作为多源、多汇节点以提高学习有效性。在每个平滑区域内分别选取一个样本作为未标记样本,以保证携带丰富的图像特征信息。利用标记样本和无标记样本作为原图像的数据子集构图,以减小构图规模进而为分类大规模图像提供条件。实验结果表明,该算法在降低时间复杂度的同时提高了图像分类的正确率,适用于大规模特征复杂的图像分类。

关键词: p电压, 电网络系统, 半监督, 图像分类, 均值漂移

CLC Number: