CAJ | 학술논문

针对基于SIFT特征的商标检索其计算量大、检索速度慢、效率低的不足，提出了一种融合小波变换和SIFT特征的商标检索方法。该方法先通过小波分解得到低频子图像并建立低频子图像库，在低频子图像上提取商标图像的SIFT特征，再逐一进行精确匹配，对相似度进行排序，输出检索结果。实验结果表明，该方法不仅保持了SIFT特征的良好描述能力，且减少了精确匹配的计算量，提高了匹配速度。融合小波变换和SIFT特征的商标检索方法有效地弥补了原算法的不足，表现出了更好的检索能力和性能。
침대기우SIFT특정적상표검색기계산량대、검색속도만、효솔저적불족，제출료일충융합소파변환화SIFT특정적상표검색방법。해방법선통과소파분해득도저빈자도상병건립저빈자도상고，재저빈자도상상제취상표도상적SIFT특정，재축일진행정학필배，대상사도진행배서，수출검색결과。실험결과표명，해방법불부보지료SIFT특정적량호묘술능력，차감소료정학필배적계산량，제고료필배속도。융합소파변환화SIFT특정적상표검색방법유효지미보료원산법적불족，표현출료경호적검색능력화성능。
Trademark search based on SIFT has some shortcomings,such as the large amount of calculation,slow search speed,low effi-ciency and so on. A method for trademark search by integration of wavelet transform and SIFT features was proposed. First,get low-fre-quency sub-image by wavelet decomposition,and create a low-frequency sub-image database. Then,extract SIFT feature in the low-fre-quency sub-image after wavelet transform,do exact match one by one,and be sort of similarity,thus output search results. The experi-mental results show that this method not only maintains a good description of SIFT features,but also reduces the amount of calculation of precisely matching and improves matching speed. The trademark retrieval method combining wavelet transform and SIFT features can ef-fectively compensate for the deficiencies of the original algorithm,and show better search capabilities and performance.