Feature Tracking in Time-Varying Volumetric Data through Scale Invariant Feature Transform

Abstract

Recent advances in medical imaging technology enable dynamic acquisitions of objects under movement. The acquired dynamic data has shown to be useful in different application scenarios. However, the vast amount of timevarying data put a great demand on robust and efficient algorithms for extracting and interpreting the underlying information. In this paper, we present a gpu-based approach for feature tracking in time-varying volumetric data set based on the Scale Invariant Feature Transform (SIFT) algorithm. Besides, the improved performance, this enables us to robustly and efficiently track features of interest in the volumetric data over the time domain. As a result, the proposed approach can serve as a foundation for more advanced analysis on the features of interest in dynamic data sets. We demonstrate our approach using a time-varying data set for the analysis of internal motion of breathing lungs.