摘要
针对图像检索存在性能的不稳定性、相对平移、旋转和尺度变换的鲁棒性不好等问题,提出了一种新的基于形状的图像检索算法.此算法通过提取轮廓的7个矩不变量和轮廓力矩来描述图像特征,用图像特征向量的归一化欧氏距离表示图像的相似度,并用星星形状的图像作为示例图像进行验证,实验证明该算法具有一定的优越性。
Two main drawbacks are identified in the recent shape-based image retrieval systems, the first is performance instability, and the other is the insufficient robustness with respect to motion, rotation and scaling. To redress these problems, a new shape-based image retrieval algorithm is proposed in this paper. The algorithm employs a set of seven moment invariants and the contour moment to extract the features of image. To describe the similarity, the Euclidean distance between two images＇ normalized vectors is used. The algorithm is tested with the given image of fires of heaven, and the experimental results indicate the certain advantages of the proposed algorithm.
出处
《宁波大学学报（理工版）》
CAS
2007年第4期438-440,共3页
Journal of Ningbo University：Natural Science and Engineering Edition
关键词
图像检索
图像形状
矩不变量
轮廓力矩
image retrieval
image shape
moment invariant
contour moment
