一种具有灰度约束的二维Otsu分割方法研究

摘要

当图像中目标类类内方差与背景类类内方差差异较大时，灰度约束Otsu分割法能够获得更好的分割结果。然而，灰度约束Otsu分割法降低了分割阈值，因而更容易受到图像中噪声的影响而出现分割错误。针对这一问题，本文将灰度约束策略推广到了二维Otsu分割法，从而提出了一种具有灰度约束的二维Otsu图像分割方法，即根据二维Otsu分割法的阈值特点选定二维灰度约束值并获得灰度约束二维直方图，而后再使用灰度约束二维直方图选取分割阈值。实验结果表明，本文方法能够更为有效的抵抗图像中的噪声，可以获得更好的图像分割结果。%When the variance of foreground and background is big, better segmentation results can be obtained by gray constrained Otsu method. However, as the threshold value of gray constrained Otsu method is smaller than Otsu method, it is more vulnerable to the influence of noise pixels in the image, which will cause segmentation errors. To solve this problem, the gray constraint strategy is extended to two dimensional histogram , which is proposed as gray constrained two dimensional Otsu segmentation. That is, selecting the two dimensional gray constrained value according to the characteristics of two dimensional Otsu method, then calculate gray constrained two dimensional histogram and get the segmentation threshold using the gray constrained two dimensional histogram. Experimental results show that, segmentation results of the proposed gray constrained two dimensional Otsu segmentation method outperform the gray constrained Otsu segmentation method when there are noise pixels in images.