Image foresting transform: on-the-fly computation of segmentation boundaries
SCIA'11 Proceedings of the 17th Scandinavian conference on Image analysis
Hi-index | 0.00 |
This work presents two fast and iterative methods that integrate segmentation by watershed and three-dimensional visualization, while the classical approach is to separate these two processes. The user-aided segmentation is based on iterative watershed, efficiently implemented using the Image Foresting Transform (IFT). The first proposed algorithm consists of extracting segmented structures' borders during the segmentation and updating them at each iteration. Structure visualization is achieved by border projection. The second algorithm updates the image directly from the changes of the segmented scene occured in each step. The two methods were implemented in C and tested with various Magnetic Resonance Images. The results are satisfactory as compared to the classical approaches. The incremental image algorithm is a little faster than the classical one while the incremental border algorithm, although a little slower in segmenting, allows a very fast rendering, very suitable for object manipulation. Thus, the proposed methods provide a good feedback to user.