基于2维灰度熵阈值选取快速迭代的图像分割

Image segmentation based on the fast iteration for two-dimensional gray entropy threshold selection

目的为了使图像阈值分割的精度和速度进一步提高,提出了一种基于2维灰度熵阈值选取快速迭代的图像分割方法。方法首先,提出了1维灰度熵阈值选取的快速迭代算法;然后,考虑图像目标和背景的类内灰度均匀性,给出了基于灰度—邻域平均灰度级直方图的灰度熵阈值选取准则;最后,提出了2维灰度熵阈值选取的快速迭代算法,并采用递推方式计算准则函数中的中间变量,避免其重复运算,加快了运算速度,大大减少了运算量。结果大量实验结果表明,与近年来提出的3种阈值分割法相比,所提出的方法分割性能更优,分割后的图像中目标区域完整,边缘清晰,细节丰富且运行时间短,仅为基于混沌小生境粒子群优化的二维斜分倒数熵分割法运行时间的3%左右。结论本文方法对不同类型灰度级图像的分割效果及运行速度均有明显优势,是实际系统中可选择的一种快速有效的图像分割方法。

Objective To further improve the accuracy and speed of image threshold segmentation, an image segmentation method is proposed based on fast iteration for two-dimensional gray entropy threshold selection. Method First, a fast itera-tive algorithm for threshold selection that uses one-dimensional gray entropy is proposed. Gray level uniformity within the object cluster and background cluster is then considered, and two-dimensional gray entropy criterion for threshold selection based on gray level-neighborhood average gray level histogram is presented. Finally, a fast iterative algorithm for threshold selection that uses two-dimensional gray entropy is proposed. In addition, recursive algorithms are adopted to calculate the intermediate variables involved in criterion function, thereby avoiding their repetitive computation. Thus, calculating speed is accelerated and calculation amount is greatly reduced. Result A large number of experimental results show that, com- pared with three threshold segmentation methods, which have been recently presented, the proposed method has superior image segmentation performance. In the segmented image, object region is complete, edges are clear, and details are rich. Moreover, running time is short and is only approximately 3% of the running time of reciprocal entropy thresholding method with two-dimensional histogram oblique division based on niche chaotic mutation particle swarm optimization. Conclusion The proposed method has obvious advantages in segmentation results and algorithmic running speed for various gray level images. It is a fast and effective segmentation method that can be used in practical systems.