地面复杂背景下自适应聚类分割方法

THE ADAPTIVE METHOD OF CLUSTERING SEGMENTATIONIN THE COMPLEX GROUND BACKGROUND

文中研究了一些自动门限法的特点，分析其对复杂背景下低反差的小目标图像分割失败的原因，提出了一种新的选取局部门限的方法。该方法先对整幅图像用聚类准则分割，然后计算分割得到的区域的信息量，看是否满足定义的门限，对不满足的区域继续分割，直到所有的被分割区域都满足条件或是分割结果不再变化。实验结果表明，该方法具有从低反差图像中抽取小目标的优良性能。

In this paper, the characteristics of some methods for the automatic thresholding are in- vestigated, and the reason they fail in the low-contrast image of complex background is analyzed . A new method of acquiring the local threshold is proposed, it segments the image with clustering method firstly, then judges the resultant region from the segmentation by calculating the information value, continues to segment the region that doesn ' t satisfy the defined threshold until all of the regions fit the threshold or the result doesn ' t change. The results of experiment show this method has good property of extracting the small target from the low-contrast image .