Wang Yugang, Huang Ting-Zhu, Wang Hui
J Opt Soc Am A Opt Image Sci Vis. 2015 Nov 1;32(11):2237-46. doi: 10.1364/JOSAA.32.002237.
In this paper, by employing the cosine function to express the so-called data fitting term in traditional active contour models, we propose an active contour model with the global cosine fitting energy for segmenting synthetic and real-world images. After that, in order to segment the image with intensity inhomogeneity, we extend the proposed global model to the local cosine fitting energy. In addition, we introduce level set regularization terms into the proposed models to avoid the expensive computational cost which is usually caused by the reinitialization of the evolving level set function. Experimental results indicate that the proposed models are accurate and effective when applied to segment different types of images. Moreover, our models are more efficient and robust for segmenting the images with strong noise and clutter than the Chan-Vese model and the local binary fitting model.
在本文中，通过使用余弦函数来表示传统活动轮廓模型中所谓的数据拟合项，我们提出了一种具有全局余弦拟合能量的活动轮廓模型，用于分割合成图像和真实世界图像。在此之后，为了分割具有强度不均匀性的图像，我们将所提出的全局模型扩展为局部余弦拟合能量。此外，我们将水平集正则化项引入到所提出的模型中，以避免通常由演化水平集函数的重新初始化所导致的高昂计算成本。实验结果表明，所提出的模型在应用于分割不同类型的图像时准确且有效。此外，对于分割具有强噪声和杂波的图像，我们的模型比Chan-Vese模型和局部二值拟合模型更高效且更稳健。
