...
  • 主题
高级搜索  >
历史搜索 清空历史
首页> 外文期刊>Journal of Medical Imaging and Health Informatics >A Novel Sketch-Based Medical 3D Model Retrieval Approach by Multi-Scale Weighted Gabor Feature Fusion and IP-HOG Feature Extraction
【24h】

A Novel Sketch-Based Medical 3D Model Retrieval Approach by Multi-Scale Weighted Gabor Feature Fusion and IP-HOG Feature Extraction

机译：通过多尺寸加权Gabor特征融合和IP-Hog特征提取的新型草图的医学3D模型检索方法

获取原文
获取原文并翻译 | 示例
           
页面导航
  • 摘要
  • 著录项
  • 相似文献
  • 相关主题

摘要

Medical 3D model retrieval plays an important role in the diagnosis by doctors. When the 3D models are used to search by doctors, it is not easy to describe the characteristics of 3D models. Therefore, using freehand sketches as inputs is a research pattern to promote medical 3D model retrieval. In the process of similarity comparison between freehand sketches and 3D models, firstly, the 3D models need to be projected into 2D images. This paper proposes an improved 2D projection algorithm for medical 3D models, which is projected into 22 2D images by sphere projection, including 6 orthographic projection images, 8 isometric projection images and 8 spherical center projection images. Secondly, extracting the features of freehand sketches and 2D projection images. A multi-scale weighted Gabor feature fusion algorithm is proposed, which can not only obtain the multi-scale and multi-direction features of medical images, but also solves the difficulties of oversize image dimension and data through the process of weighted Gabor feature fusion. Finally, on the basis of multiscale weighted Gabor feature fusion images, a histogram of oriented gradient based on interest points (IP-HOG) feature extraction algorithm is proposed. The IP-HOG feature extraction algorithm can effectively extract the local features of each block, which is according to the number of interest points of different blocks in multi-scale weighted Gabor feature fusion images. The experimental results show that our 2D projection algorithm achieve better retrieval performance, which has an average increase in precision value is around 15.3%: our multiscale weighted Gabor feature fusion algorithm combined with IP-HOG feature extraction algorithm for feature extraction, which has an average increase in precision value is around 10.3%. Compare with several leading sketch-based 3D model retrieval approaches, our approach has an average increase in precision value is around 13.1%.
机译：医学3D模型检索在医生的诊断中起着重要作用。当3D模型用于被医生搜索时，描述3D模型的特性并不容易。因此，使用手绘草图作为输入是促进医学3D模型检索的研究模式。在手法草图和3D模型之间的相似性比较过程中，首先，需要将3D模型投影到2D图像中。本文提出了一种改进的医学3D模型投影算法，其通过球体投影投射到22个2D图像中，包括6个正射投影图像，8个等距投影图像和8个球形中心投影图像。其次，提取手绘草图和2D投影图像的特征。提出了一种多级加权Gabor特征融合算法，其不仅可以获得医学图像的多尺度和多向特征，而且还通过加权Gabor特征融合的过程解决了超大图像维度和数据的困难。最后，在多尺度加权Gabor特征融合图像的基础上，提出了一种基于兴趣点（IP-Hog）特征提取算法的面向梯度的直方图。 IP-Hog特征提取算法可以有效地提取每个块的本地特征，这是根据多尺度加权Gabor特征融合图像中不同块的感兴趣点数的局部特征。实验结果表明，我们的2D投影算法实现了更好的检索性能，该性能平均增加约为15.3％：我们的MultiScale加权Gabor特征融合算法与IP-Hog特征提取算法相结合，具有平均值精度增加约为10.3％。与基于几种主要的草图3D模型检索方法进行比较，我们的方法平均增加了精度值约为13.1％。

著录项

相似文献

  • 外文文献
  • 中文文献
  • 专利
获取原文

客服邮箱：kefu@zhangqiaokeyan.com

京公网安备：11010802029741号 ICP备案号：京ICP备15016152号-6 六维联合信息科技 (北京) 有限公司©版权所有

  • 客服微信

  • 服务号