基于独立关键子块和三角树的快速图象检索新方法

由于图象存储数据量非常大,因此提取图象特征和检索极为耗时.为了提高图象检索效率,将文本检索中的有效检索方法(基于关键字频率与关键字逆文档频率乘积的索引模型)结合三角树索引机制应用到基于内容的图象检索,提出了一种基于独立关键子块和三角树的快速图象检索新方法.该方法首先用独立分量分析将样本图象子块中的直方图特征映射到色彩概念空间来得到类似于文本中关键字的独立关键子块;然后再用训练好的模糊支持向量机去识别每幅图象中所包含的独立关键子块,由于独立分量分析能够使特征彼此保持高阶独立性,因此该方法与主成分分析方法对比,具有较高检索效率;最后,再通过构造三角树来来为图象数据库建立分层索引结构,以加快检索速度.

Fast Image Retrieval Method based on Independent Keyblock and Triangle Tree

Because image database is very huge, the feature extraction and retrieval process are usually time consuming. In order to effectively use existing text information retrieval methods in content based image retrieval, especially the index mechanism of the product tf * idf by term frequency (tf) and inverse document frequency (idf) for each text document, this paper cooperates tf * idf model with triangle tree to improve the retrieval performance. First, after pixel-based histogram features of sub-block in certain image class are mapped to color concept space through independent component analysis (ICA), we would obtain all of independent keyblock of such image class; then well-trained fuzzy support vector machine is used to recognize all of independent keyblocks contained by each image. Similar to text retrieval, in which the whole text document is indexed by , the recognized independent keyblock is used to index each image in database. Because independent component features are naturally high order independent with each other, compared to principle component analysis (PCA) method, this algorithm achieves higher performance. At last, triangle tree is used to hierachically index image database and thereof speed up retrieval.