Improved Live-Wire algorithm for kidney image segmentation

摘要

Live-Wire segmentation algorithm is an interactive tool for efficient, accurate and reproducible boundary extraction which requires minimal user input. But the traditional Live-Wire algorithm is sensitive to noise and inefficient to distinguish between strong and weak edges in the image. To overcome the two drawbacks, this paper proposes an improved Live-Wire algorithm, its cost function is redefined: (1) change the Laplace operator into Canny and Sobel operators to construct the multiple cost terms, differences between Canny and Sobel operators allow them to contribute to the total cost calculation without redundant information; (2) redefine the gradient magnitude function of the cost function as nonlinear weighting edge gradient, the nonlinear weighting procedure can solve the undistinguished edge problem very well. Experiments are conducted on simulation and real image of the kidney. The experimental results prove that the proposed algorithm is really of good performance in the two drawbacks.