Local region-scalable active contour using expandable kernel

Abstract

This thesis presents a novel active contour using scalable local regional information on expandable kernel for image segmentation. We call it LREK active contour. Our model uses intensity values of pixels on a set of scalable kernels along evolving contour. These kernels are to direct contour front towards object’s boundary within an image domain. Key feature of our model is that scale of the kernels increases gradually until the boundary is detected. So, our LREK may reach the boundary faster than some other methods. We compare performance of our LREK to existing edge and region-based active contour models. Experimental results show more desirable segmentation outcomes of our method. Furthermore, we also extract directional property of scalable local regional information so that it can choose objects of desirable edge’s type. In addition to an ability in segmenting two different edge’s type objects with only one initial contour, our proposed scheme performs effectively in segmenting noisy, concave boundary, non-uniform, and heterogeneous textures objects with a large capture range and fast convergence. Meanwhile, level set formulation makes our model topologically flexible. Moreover, our Gaussian LREK is able to trace blur or smooth boundary.