基于兴趣点凸包的图像检索方法

New image retrieval method based on convex hull of interest points

提出了一种基于兴趣点确定感兴趣区域的图像检索方法。在尺度归一化图像中检测兴趣点,然后对兴趣点求取凸包确定感兴趣区域,并用颜色直方图和Zernike矩作为特征进行检索,在保证图像检索对图像旋转、平移、缩放鲁棒性的前提下,将图像检索上升到物体层,克服了传统方法的不足。对1000幅图像进行的大量实验表明,该方法与其他基于兴趣点的方法相比,平均检索准确率提高了13%,可以更准确地查找到用户所需图像。

A new image retrieval method based on region of interest determined by interest points is presented.Interest points are detected from the scale normalized image.Then,the convex hull of interest points is calculated to extract the region of interest in the image.Color histogram and Zernike moments are used as image features to carry out retrieval.With robustness to image rotation,translation and scale,the method makes the retrieval implementation on the object level and avoids the shortcoming of traditional methods.Lots of experiments on 1000 images show that the method improves the average retrieval precision by 13%,compared with other interest points based retrieval methods.