基于最大熵-方差模型的图像分割方法

Threshold Image Segmentation Based on Maximum Entropy-Variance Model

针对当图像中目标与背景的面积相差很大时,最大类间方差方法的分割性能迅速下降的问题,研究了信息熵和方差的关系。认为信息熵和方差都被用作不确定性的度量,两者之间定会存在一定的科学关系,因此将最大熵和最大类间方差结合起来建立数学模型,提出基于最大熵-方差模型的图像分割方法,并引入类内方差对分割进行评价来选取参数调整算法的分割性能,更充分地利用了图像的灰度信息。通过实验证明该方法优于最大熵方法和最大类间方差方法,具有较强的稳定性,提高了图像分割精度。

When the area of the target and the background for an image are more different,the performance to segment an image by the maximum between-class variance method declines rapidly.So have researched the relation between information entropy and variance,think that both information entropy and variance are used as a measure of uncertainty,there must be certain scientific relations between the two.Has constructed the mathematical model between maximum entropy and maximum variance,a new image segmentation method based on entropy-variance model is proposed.Adjust the performance to segment an image through evaluating the variance within the class,make better use of the gray information of an image.Numerous experiments show that the method is better than the method of maximum entropy and the maximum between-class variance method,the method is very stable and enhance the accuracy of image segmentation.