Multiple active contours using scalable local regional information on expandable kernel

In this paper, we present multiple active contours for image segmentation that use scalable local regional information on expandable kernel. It includes using a strategy inside the variational level set method to adapt the size of a local window in order to avoid being stuck locally in a homogeneous region during the segmentation process. It also provides a multiple level set framework to deal with simultaneous multiple object segmentation without merging and overlapping between adjacent contours in the shared boundaries of separate regions. Several experiments are demonstrated to validate its segmentation performance. It shows that the neighbouring contours can avoid merging and overlapping in the shared boundaries that has low contrast. In segmenting multiple regions, the choice of various parameters can be different for each zero level contour which contributes to accuracy improvement in the segmentation outcomes.