A method of shape encoding and retrieval is proposed in this letter, which uses centripetal code to encode shape and extracts shape's convex for retrieval. For the rotation invariance and translation invariance of...A method of shape encoding and retrieval is proposed in this letter, which uses centripetal code to encode shape and extracts shape's convex for retrieval. For the rotation invariance and translation invariance of the centripetal code and the normalization of convex,the proposed retrieval method is similarity transform resistant, Experimental results confirm this capability.展开更多
文摘为提高图像形状信息的检索准确率和效率,提出应用全方向形状特征码的图像检索方法.该方法在全方向上对形状进行分割,度量形状各方向各部分的复杂度,构建形状的全方向特征码,计算形状间的相似度.通过真实建筑形状数据集和MPEG-7 CE-1 Part B形状数据集对本方法进行了检索性能测试,并和其他形状相似性描述方法进行了对比.实验结果表明,本方法具有更高的检索准确率和较高的检索效率.全方向形状特征码图像检索方法可以准确描述形状的全局特征与局部特征,具有平移、旋转、尺度不变性,及较强的形状描述识别能力.
基金National Natural Science Foundation of China(No. 60172045)863-306 Project (863-306-ZT03-09)
文摘A method of shape encoding and retrieval is proposed in this letter, which uses centripetal code to encode shape and extracts shape's convex for retrieval. For the rotation invariance and translation invariance of the centripetal code and the normalization of convex,the proposed retrieval method is similarity transform resistant, Experimental results confirm this capability.