摘要
提出了一种新的单基因相位描述子的自然图像检索方法,首先对图像进行Riesz变换,获得图像的能量信息和结构信息,然后编码局部区域的单基因相位变化和每个像素的单基因相位特征,接着计算抽取的局部特征的统计特征并运用基于块的fisher判别分析进行特征降维,最后进行相似性度量.实验结果表明提出的方法能够有效地描述自然图像,明显改善了图像检索的精度.
A novel method for content-based image retrieval based on monogenic phase descriptor is proposed, firstly, Ricsz transform is performed to an original image, and the energetic and structure information of the image are obtained, then the monogenic phase variation in each local region and monogenic phase feature in each pixel are encoded, and next the statistical features of the extracted local features are calculated and the dimensionality of the local statistical feature is reduced using block-based fisher discriminant analysis, finally, similarity measurement between query image and any image of database is computed. Experimental results show that the proposed method can effectively describe an image, and obviously improve the average retrieval precision.
出处
《福建师范大学学报(自然科学版)》
CAS
CSCD
北大核心
2014年第4期20-25,31,共7页
Journal of Fujian Normal University:Natural Science Edition
基金
福建省科技厅重点资助项目(2013H0020)
关键词
图像检索
单基因表示
局部特征
基于块的fisher判别分析
相似性度量
image retrieval
monogenic representation
local feature
block-based fisher discriminant analysis
similarity measurement.