CAJ | 학술논문

传统基于局部特征表示的图像检索方法在图像特征提取和特征相似性匹配时计算量较大，为此提出一种运用随机算法进行改进的图像检索方法。在图像特征提取方面，通过随机采样获得数量适当的像素点作为特征点，用SIFT（scale invariant feature transform）算子对随机特征点进行描述以形成图像的有效表示；在特征相似性匹配方面，采用基于随机映射的LSH（locality sensitive hashing）算法为图像特征库建立索引，并用于对所查询图像的局部特征进行高效的近似近邻搜索。实验结果表明，该方法有效降低了图像检索的计算复杂度，提高了检索效率。
전통기우국부특정표시적도상검색방법재도상특정제취화특정상사성필배시계산량교대，위차제출일충운용수궤산법진행개진적도상검색방법。재도상특정제취방면，통과수궤채양획득수량괄당적상소점작위특정점，용SIFT（scale invariant feature transform）산자대수궤특정점진행묘술이형성도상적유효표시；재특정상사성필배방면，채용기우수궤영사적LSH（locality sensitive hashing）산법위도상특정고건립색인，병용우대소사순도상적국부특정진행고효적근사근린수색。실험결과표명，해방법유효강저료도상검색적계산복잡도，제고료검색효솔。
An image retrieval method using random algorithms is proposed to improve the traditional local feature representation method which often needs a large amount of calculation during image fea‐ture extraction and similarity matching .For image feature extracting ,the method adopts random sam‐pling to obtain an appropriate number of image pixels as the feature points ,then represents these ran‐dom feature points with SIFT descriptors in order to form an effective image representation .For fea‐ture similarity matching ,it applies a random mapping LSH algorithm to indexing the feature database and conducting the efficient approximate nearest neighbor query of image local features .Experimental results show that the proposed method can efficiently reduce the computation complexity and improve the image retrieval efficiency .