结合图像过渡区的变分水平集模型

VARIATIONAL LEVEL SET MODEL BASED ON IMAGE TRANSITION REGION

为解决无需重新初始化水平集演化模型对初始轮廓敏感的问题,基于过渡区提出一个变分水平集模型。首先从形态学角度提取图像的过渡区,从而获得分割图像的阈值;然后根据这个阈值,加权面积项前的尺度参数修改为一个函数,它在感兴趣的目标内外有相反符号,水平集函数可初始化为一个常值函数。提出的模型不但从根本上解决了对初始轮廓敏感的问题,而且能够实现对图像的快速分割。对合成和真实的图像的实验验证了该模型的有效性。

Level set evolution without re-initialisation （LSWR） model is sensitive to initial contour. In order to solve this problem, a variational level set model based on image transition region is proposed. First, it extracts the transition region of image in perspective of morphology and thereby obtains the threshold of segmentation; Then it modifies the scale parameter associated with weighted area as a function according to this threshold, which is shown to have opposite signs inside and outside the interested objects, therefore the level set function can be initialised to a constant function. The proposed model resolves fundamentally the problem of sensitive to initial contour, moreover, it can also quickly realise the fast segmentation of images. Experiments on synthetic and real digital images verify the effectiveness of the model.