Hide/Show Apps

Computationally efficient shape analysis via level sets

Shah, Jayant
Pien, Homer
In recent years, curve evolution has been applied to smoothing of shapes and shape analysis with considerable success, especially in biomedical image analysis. The multiscale analysis provides information regarding parts of shapes, their axes or centers and shape skeletons. In this paper, we show that the level sets of an edge-strength function provide essentially the same shape analysis as provided by curve evolution. The new method has several advantages over the method of curve evolution. Since the governing equation is linear, the implementation is simpler and faster. The same equation applies to problems of higher dimension. An important advantage is that unlike the method of curve evolution, the new method is applicable to shapes which may have junctions such as triple points. The edge-strength may be calculated from raw images without first extracting the shape outline. Thus the method can be applied to raw images. The method provides a way to approach the segmentation problem and shape analysis within a common integrated framework.