Abstract
This paper proposes active contour models that splitting the control points can extract precisely and effectively an apex part from the extraction region in images. The problem of active contour models is that the contour of a smooth shape is converged relatively correctly, but that of a corner such as an apex is hard to converge due to the characteristics of these models. To solve this problem, some technical methods using shape models were suggested. However, such methods were under the restrictions that the contour of a shape in the extraction region must be similar to a shape model and prior knowledge of an extraction object must be required. The technical method presented in this study is to detect a high stress point between the control points in the contour convergence process of active contour models and to split the corresponding control points. As a result, the degree of freedom of a split control point increases and the control points are located closer to each other at an apex part than a smooth shaped part in the extraction region, which allows an apex part to be extracted precisely and effectively from the whole extraction region in images. We applied this method to an artificial image, conducted its experimental evaluation and proved its effectiveness.
