关键词: 基于内容的图像检索, 商标, 傅里叶描述子, 尺度不变特征转换, 全局特征, 局部特征

Abstract: Traditional trademark image retrieval algorithm only using global feature easily make mistaken retrieval, and Scale Invariant Feature Transform (SIFT) features have limited descriptive ability for image contour and high algorithm complexity. According to the problem above, a new retrieval algorithm for trademark was proposed which combined the global feature and the local feature of images. The global feature reflects the overall information of the image that can help to build the candidate image database quickly, while the local feature can be matched with the candidate images more accurately. Firstly, extract Fourier Descriptor (FD) of the retrieved image and sort them according to similarity. And then, based on this result, the candidate images were matched accurately through extracting the SIFT features. The experimental results show that this method not only keeps high recall-precision of SIFT features and is superior to the method based on the single FDs feature, but also compared to the single SIFT features it effectively improves retrieval speed. This method can be well applied in the trademark image retrieval system.

Key words: Content-based Image Retrieval (CBIR), trademark, fourier descriptors, Scale Invariant Feature Transform (SIFT), global feature, local feature