2004 年 24 巻 Supplement1 号 p. 453-454
This paper focuses on volume data generated from medical CT/MRI images and provides a method for indexing and retrieval of volume data sets based on their 3D shape. The proposed method indexes volume data by calculating the distribution of its critical points, where the gradient is zero. Sine the volume data sets generated from medical CT/MRI images have a lot of critical points derived from measurement noise, a critical-point-based metric between volume data is obtained at high computational cost, and the size of indexing data is large. In order to reduce the number of critical points without loss of geometric features, Morse theory is applied. Retrieval examples show the efficiency of the proposed method.