Wavelet-based multiscale DDGVF snake model for cardiac magnetic resonance images segmentation

Active contour model is the genaral method used for left ventricle segmentation in cardiac magnetic resonance images. Because of the sensitivity to initial position for active contour, unreasonable initial position will lead to unaccuracy and unstable results. In this article, the author incorporated wavelet-based multiscale edge detection and dynamic directional gradient vector flow into the active contour model. Preliminary processing was performed at high-scale image and the results were used as initial contour in low-scale image for the following processing. The problem that the initial position must be close to the actual edge was mostly solved and the capture range of the active contour was enlarged.