• 主题
高级搜索  >
历史搜索 清空历史
首页> 外文学位 >Robust computer vision algorithms for registering images from curved human retina.
【24h】

Robust computer vision algorithms for registering images from curved human retina.

机译：强大的计算机视觉算法，用于记录来自弯曲的人类视网膜的图像。

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

Accurate registration is essential for mosaic synthesis, change detection, and design of computer-aided instrumentation. A linear, feature-based, non-iterative method for jointly estimating consistent transformations of all images onto the mosaic "anchor image" is presented. Constraints for this joint estimation are derived from pairwise registration. Central to the pairwise registration algorithm is a 12-parameter inter-image transformation derived by modeling the retina as a quadratic surface with unknown parameters, assuming an uncalibrated weak perspective camera and rigid motion of the eye between images. The parameters of this model are estimated by matching vascular landmarks extracted by an algorithm that recursively traces the blood vessel structure. The parameter estimation technique is a hierarchy of models and methods: an initial match set is pruned based on a 0th order transformation estimated using a similarity-weighted histogram; a 1st order, affine transformation is estimated using the reduced match set and least-median of squares; and the final, 2nd order, 12-parameter transformation is estimated using an M-estimator initialized from the 1st order results. This hierarchy makes the algorithm robust to unmatchable image features and mismatches between features caused by large inter-frame motions. Before final convergence of the M-estimator, feature positions are refined and the correspondence set is enhanced using normalized sum-of-squared differences matching based on the emerging transformation. The constraints computed by the pairwise registration are used in the joint solution. An incremental, graph-based technique constructs the set of registered image pairs used in the joint solution; all O(N 2) image pairs need not be registered directly. The joint estimation technique allows images that do not overlap the anchor frame to be successfully mosaiced, a particularly valuable capability for mosaicing images of the retinal periphery when diagnosing and treating diseases such as AIDS/CMV Experiments applied to data sets from 16 eyes show the overall, average transformation error in final mosaic construction to be 0.80 pixels.
机译：精确配准对于镶嵌合成，变化检测和计算机辅助仪器设计至关重要。提出了一种线性，基于特征的非迭代方法，用于共同估计所有图像到镶嵌“锚图像”上的一致变换。此联合估计的约束是从成对配准导出的。成对配准算法的核心是通过将视网膜建模为具有未知参数的二次曲面而得出的12参数图像间变换，假定未校准的弱透视相机和图像之间眼睛的刚性运动。该模型的参数通过匹配递归地跟踪血管结构的算法提取的血管界标来估算。参数估计技术是模型和方法的层次结构：基于使用相似加权直方图估计的0阶变换，对初始匹配集进行修剪；使用减少的匹配集和最小二乘方估计一阶仿射变换；然后使用从一阶结果初始化的M估计器来估计最终的二阶12参数变换。这种层次结构使该算法对于不可匹配的图像特征以及由较大的帧间运动引起的特征之间的失配具有鲁棒性。在M估计量最终收敛之前，可以使用特征化位置，并使用基于新兴变换的归一化平方和差匹配来增强对应集。通过成对配准计算的约束条件用于联合解决方案。一种基于图的增量技术可构造联合解决方案中使用的一组配准图像对。无需直接注册所有O（N 2）个图像对。联合估计技术可以使不重叠锚帧的图像成功地进行镶嵌，这在诊断和治疗诸如AIDS / CMV的疾病时对镶嵌视网膜周边图像具有特别有价值的功能，将其应用于16只眼睛的数据集上的实验显示出整体，最终镶嵌构造中的平均变换误差为0.80像素。

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