关键词: 主聚类, 图像检索, 轮廓曲线

Abstract: Traditional Content-Based Image Retrieval（CBIR） and tracking algorithm mainly uses image color，texture and other features as similarity comparison between two images.However，a large number of experiments and applications also show that it is difficult to precisely control spatial structure and object shape with color and texture for images similarity comparison，and unexpected results are often produced during image retrieving.In order to enhance precision for image retrieval，an image retrieval method containing features of color and object contour curve is presented.Image is segmented and the contour of interested object in image is extracted，and then the contour is transformed by affine and is processed by the minimum.The contour contains the whole information of interested object，and preserves the geometric invariance;with color feature，a histogram for primary cluster is extracted.The histogram extracted contains not only the color information but also space location information for primary cluster.The weighted average for color distance histogram and distance deviation of contour curve is applied as similarity measure between two images.Experiments show that the presented algorithm obtains more robust retrieval precision.

Key words: primary cluster, image retrieval, contour curve