山东师范大学学报(自然科学版)
山東師範大學學報(自然科學版)
산동사범대학학보(자연과학판)
JOURNAL OF SHANGOND NORMAL UNIVERSITY(NATURAL SCIENCE)
2014年
4期
14-17
,共4页
图像检索%动态特征向量%特征提取%模糊量化
圖像檢索%動態特徵嚮量%特徵提取%模糊量化
도상검색%동태특정향량%특정제취%모호양화
image retrieval%dynamic characteristic vector%feature extraction%fuzzy quantization
针对传统特征融合方法中权值的不易确定性,提出一种动态确定特征向量权值的算法。首先根据特征向量对应维之间的距离确定权值大小并形成权值矩阵,然后计算加权特征向量之间的距离,作为两幅图像的相似度。最后通过相似度排序完成图像检索。实验数据表明,该算法在存在干扰数据集上都有良好的检索效果,由于不需要大量实验确定权值的大小,在效率上得到了大大提高,与传统的算法相比,具有更高的检索效率。
針對傳統特徵融閤方法中權值的不易確定性,提齣一種動態確定特徵嚮量權值的算法。首先根據特徵嚮量對應維之間的距離確定權值大小併形成權值矩陣,然後計算加權特徵嚮量之間的距離,作為兩幅圖像的相似度。最後通過相似度排序完成圖像檢索。實驗數據錶明,該算法在存在榦擾數據集上都有良好的檢索效果,由于不需要大量實驗確定權值的大小,在效率上得到瞭大大提高,與傳統的算法相比,具有更高的檢索效率。
침대전통특정융합방법중권치적불역학정성,제출일충동태학정특정향량권치적산법。수선근거특정향량대응유지간적거리학정권치대소병형성권치구진,연후계산가권특정향량지간적거리,작위량폭도상적상사도。최후통과상사도배서완성도상검색。실험수거표명,해산법재존재간우수거집상도유량호적검색효과,유우불수요대량실험학정권치적대소,재효솔상득도료대대제고,여전통적산법상비,구유경고적검색효솔。
Considering the difficulty in deciding the feature vector in the traditional feature fusion method,we provided a method to dynamically decide the feature vector of the eigenvectors.Firstly,the feature vector is decided by calculating the distance between eigenvectors,and the feature vector matrix can be obtained.Then,the similarity degree of two pictures can be measured by the distances of the normalized eigenvectors.Finally,the searching data are ordered according to the similarity degree with the inputting data,and the whole searching process is completed after this operation.Our experimental results show that this algorithm is quite effective in the data set where interferences exist.Since the feature vector are not decided according to experiment,the algorithm is much more efficient the traditional algorithm.