Deformable Models
Omer Demirkaya, Musa H. Asyali, Prasanna K. Sahoo · 2008
General-purpose segmentation methods are difcult to apply to medical images due to the complex nature of medical images. Complexities arise from fuzzy boundaries, missing edge features, complicated backgrounds, and inconsistent image contrast. Approaches that solely depend on the low level information, local features, will most likely fail at borders with fuzzy boundaries and/or missing edge features. Figure 8.1 demonstrates this problem on an MR image of the ascending aorta and its edges detected by the Canny operator [1]. The Canny edge detector depends on the local features of the image and is very popular in imaging eld; however, it has failed to fully detect the aorta boundaries due to poor contrast and fuzzy borders, giving spurious edges within the vessel due to turbulent ow. It will be difcult to automatically eliminate the spurious edges or to link the missing boundaries within such a complicated background where there are too many other edge features in close proximity.