摘要
局部线性嵌入算法LLE(Locally Linear Embedding)可以有效地对图像的高维特征进行降维。针对处理样本分布不均匀及近邻因子选择时会出现的问题,在对高维数据降维时,近邻点的选择采用计算测地线距离而非传统的局部欧式距离,且近邻点的个数选择进行预先优化以达到更好的降维效果。实验表明,改进后的LLE算法具有更好的分类精确度,在图像分类过程中比单纯的LLE算法具有更好的分类性能。
Locally linear embedding （ LLE ） algorithm can effectively reduce the dimension of the image with high dimension characteristics.Aiming at the problems occurred when processing the uneven samples distribution and neighbourhood factor selection, while reducing the dimensions of high-dimension data, we use the calculation of geodesic distance instead of the calculation of unconventional local Euclidean distance to select neighbour points.Moreover, in order to achieve better effect of dimension reduction, the selection of neighbour points’ number is optimised in advance.Experiments show that the improved LLE algorithm has better classification accuracy, it has better classification performance than the pure LLE algorithm in image classification process.
出处
《计算机应用与软件》
CSCD
北大核心
2014年第12期256-258,302,共4页
Computer Applications and Software
基金
江苏省高校自然科学研究项目(14KJD520003)
关键词
局部线性嵌入
图像检索
降维
近邻因子
Locally linear embedding
Image retrieval
Dimension reduction
Neighbourhood factor
