UCLA Brain Mapping Center

Publications

Elastic shape models for interpolations of curves in image sequences.

Joshi SH; Srivastava A; Mio W;
Information processing in medical imaging : proceedings of the ... conference. 2005-Dec; 19(541-52)
 
Many applications in image analysis are concerned with the temporal evolution of shapes in video sequences. In situations involving low-contrast, low-quality images, human aid is often needed to extract shapes from images. An interesting approach is to use expert help to extract shapes in certain well-separated frames, and to use automated methods to extract shapes from intermediate frames. We present a technique to interpolate between expert generated shapes. This technique preserves salient features in the interpolated shapes, and allows analysts to model a continuous evolution of shapes, instead of a coarse sampling generated by the expert. The basic idea is to establish a correspondence between points on the two end shapes, and to construct a geodesic flow on a shape space maintaining that correspondence. This technique is demonstrated using echocardiagraphic images and infrared human gait sequences.
 
PMID: 17354724    doi: 10.1109/TPAMI.2005.86
 

BMAP Author