一种使用显著封闭边界的图像检索算法

Novel image retrieval algorithm using salient closed boundary

在分析了基于边缘的图像检索方法的缺点后,提出了一种基于显著封闭边界的图像检索新算法。它使用基于图论方法的边缘比算法,从噪声图像中获取图像的显著封闭边界,然后以显著封闭边界来代表图像,每条显著封闭边界由边缘角度直方图来刻画,并以此特征进行图像间的相似性度量。由于图像的显著封闭边界代表了图像的显著结构,减少了短小边缘的噪声影响,故对图像检索十分有效。实验结果表明,本算法具有较高的检索准确率。

Having been analyzed the drawbacks of image retrieval methods based on edges, a novel image retrieval algorithm based on salient closed boundary is presented. Firstly, the ratio contour, a novel graphbased method is performed to extract salient closed boundaries from noisy images. Secondly, salient closed boundaries are selected and each salient closed boundary is described by the feature; edge angular histogram, the image is represented by the feature of salient closed boundaries. Finally, the similarities are measured by the angular histogram of the salient closed boundary. Because the salient closed boundary represent the prominent structure of image and eliminate the noise of the short edges, the algorithm is very effective for image retrieval. Experimental results demonstrate that the proposed method is more effective than the traditional algorithm.