自适应神经树在图片搜索中的应用研究

摘要

针对传统基于内容的图片搜索由于只使用到图像低层次特征信息而存在着匹配正确率较低的问题,提出一种基于自适应神经树来进行特征提取从而得到较高层次的图像特征信息的方法,用以提高相似图像检索准确率.通过为每张图像都构建一棵能表征其特征信息的神经树并对其使用霍夫曼编码,得到以向量方式表示的图像,从而使用向量间的比较来测算图像的相似程度.实验结果表明,该方法可以提高搜索的准确率.%For the problem of low accuracy in the content based image search,a method for feature extraction based on adaptive neural tree is proposed to get high level image feature information.It can improve the accuracy of image search.Based on Construct an adaptive neural tree for each image and coding the tree,it can use the vector to represent the image.Therefore,it can measure the similarity degree of the picture by the comparison between the vectors.The experimental results show that this method can greatlv improve the accuracy of the search.