Fast search for thresholds from 1D and 2D histograms by an iterative algorithm for image segmentation

Image segmentation plays a very important role in many tasks of pattern recognition, computer vision, and image and video retrieval. In applications such as real-time recognition and tracking systems, it is desirable to develop as fast segmentation algorithms as possible. Image thresholding is one of the most important techniques for image segmentation. Its goal is to automatically find thresholds from the histograms of images, and to perform image segmentation with the thresholds. This paper proposes an efficient iterative optimization approach to the search for thresholds from the 1D or 2D histograms of images. Our algorithm is also compared with two well-known and commonly-used techniques. The experimental results show that the new algorithm performs better.