Threshold Selection Using Competitive Algorithm

Automatic threshold selection is significant in image segmentation, understanding and recognition. Finding global minimum of the histogram is difficult because many algorithms fail to end in a local minimum. A new competitive algorithm is proposed to automatically select threshold based on histogram analysis. Given a random initial value, the new algorithm begins to renew the value towards global minimum without stopping at local minimum by a competitive scheme. With the new algorithm, the briefness, low computation, and stability characteristic of threshold segmentation is reserved.