基于直方图偏差约束的快速模糊C均值图像分割法

Fast Fuzzy C-Means Image Segmentation Method Based on Histogram Deviation Constraints

为了解决传统模糊C均值（FCM）聚类分割算法计算耗时的问题，提出了在直方图偏差约束条件下的快速FCM图像分割算法．通过对原始图像重新采样以减小FCM算法数据处理的数量，利用平滑后归一化直方图的距离偏差作为约束条件来计算合适的采样率，以控制重新采样产生的图像失真，得到满足正确分割所需要的阈值，并在采样率计算中采用黄金分割法搜索满足约束条件的采样率．实验结果表明，在保持传统FCM聚类算法分割效果的前提下，所提算法的分割时间分别仅为传统的FCM、二维熵、Otsu等算法的3．0%～11.2%、9.2%～30．2%和15.0%～52．0%.

In order to solve the problem that the traditional fuzzy C-means （FCM） segment algorithm is time consuming, a fast FCM algorithm based on histogram deviation constraints is proposed, in which the initial image is re-sampled to reduce the quantity of data processing, the distance deviation of normalized histogram after smoothing is utilized as a constraint condition to calculate proper sample rate so as to control the image distortion due to resample, and the required threshold satisfying the correct segmentation is obtained. The golden section searching algorithm is used to search the sample rate meeting the constraint condition. Experimental results show that the segment time of the proposed algorithm is only 3.0%-11.2%, 9.2%-30.2% and 15.0%- 52.0% of the traditional FCM, 2D entropy and Otsu algorithm respectively while keeping the same segment effect as that the traditional FCM has.