颅脑MRI图像的分水岭分割方法研究

Research on Watershed Segmentation Algorithm of Craniocerebrum MRI Image

针对颅脑MRI图像的模糊特点和实际应用的要求,提出了一种改进的分水岭算法。从图像的结构信息考虑,由于噪声产生的谷底值很小,而对应于真正的区域,每个区域的最小谷底会有一个很大的动态值,这个值与没有噪声时的谷底动态值相近。因此,只要简单地给一个阈值,就可以将那些由噪声产生的谷底滤掉,从而也就抑制了过分割问题。结果表明,该方法能够快速、准确地得到医学图像的分割结果,并且具有较强的抗噪声能力。

In light of the fuzziness of craniocerebrum MRI image and the requirement in practical application, an improved watershed algorithm is proposed. In consideration of the structure information of image, the valley-bottom value produced by noise is very small. However, the minimum valley-bottom of each area was a very big dynamic value corresponding to real area, which is close to the valley-bottom dynamic value when there is no noise. Hence, the valley-bottom produced by noise can be filtered, thus effectively restraining the over-segmentation, provided that a threshold is simply given. Experimental results show that the algorithm can quickly and accurately obtain the segmentation result of medical image, possessing a higher noise-resistant capability.