Proposal for multiple methods of active net model

Abstract

In this paper, we propose a new method for active net model. The active net is a lattice network model that is based on minimizing energy to extract the region of a target object. In a conventional method, only one target is detected by one net. Therefore, assuming that the image can be divided into some parts, the proposed method uses many nets and applies them to each divided image. Some simulation results are presented obtained by applying the proposed method to images. In the proposed method, since the size of a net becomes small, the processing time to one net can be shortened. Furthermore, since the distance between lattice points is short in the initial net, the target at the end of an image becomes detectable. That is, it means that the target detection performance has improved.