摘要
针对图像检索存在性能的不稳定性、相对平移、旋转和尺度变换等问题,提出了基于区域内形状特征的不变矩和轮廓力矩法和傅里叶描述符结合的方法。其中的不变矩和轮廓力矩法具有良好的平移、旋转、尺度缩放不变性及抗干扰性,傅里叶算法不仅对噪音具有很好的鲁棒性,而且对几何变换具有不变性,更加适合图像检索的需要。通过实验可知,该算法对于图像的扭曲形变具有不变性,在具有一定形变干扰的情况下,仍得出较好的图像检索结果;且检索结果排列的顺序与人的主观视觉判断大致相同,检索精度好。
A new image retrieval algorithm is proposed using the same moment of the shape within the region and the outline of the moment combined with Fourier descriptors to solve the problem of instability, the relative translation, rotation and scale transformation and so on. The same moment and the outline of the moment method have good translation, rotation, scale invariant and anti-scaling. The Fourier algorithm not only has a very good robustness to noise, but also transforms geometric invariant with a more suitable image retrieval. The experiment shows that the algorithm has deformation invariance for image distortions and that the order of retrieval results agree approximately with human subjective judgments.
出处
《电子科技》
2008年第12期69-71,共3页
Electronic Science and Technology
关键词
图像检索
不变矩
轮廓力矩
相似度
image retrieval
the same moment
outline of the moment
similarity
