The segmentation of intensity inhomogeneity images is always a challenging problem. There are two kinds of intensity inhomogeneities, one associated with the imaging devices and illumination variations, and the other associated with the essential characteristics of the intensities in objects and backgrounds. We name the second kind of intensity inhomogeneity as intensity complexity. In this paper, we focus on the segmentation of the images with intensity complexity. Our main argument is to quantify the complex intensities and convert them into useful features to improve segmentation accuracy. Two new quantities called the region intensity complexity index (RIC-Index) and factor (RIC-Factor) are introduced to quantify the intensity complexity. Then the quantified intensity complexity is incorporated into a variational level set framework. The total energy functional of the proposed framework consists of the following three items: a region intensity complexity term, a local region fitting energy term, and an edge-based energy term. The first term is defined by exploiting the region intensity complexity factor of the images. Mean and variance are utilized in the local region fitting energy to describe the image texture information. The last term of the energy functional, which is also derived from the region intensity complexity factor, incorporates the significant edge information. By integrating these three terms, the proposed model can handle intensity complexity images, especially two kinds of images: one with complex intensities in the objects, and the other with complex intensities in the backgrounds. The experimental results on 40 intensity complexity images and 1000 natural images from the Extended Complex Scene Saliency Dataset have indicated that our proposed algorithm can produce satisfactory segmentation results in comparison with five state-of-the-art methods and a deep learning approach.

中文翻译：

---

区域强度复杂度活动轮廓

强度不均匀图像的分割一直是一个具有挑战性的问题。有两种强度不均匀性，一种与成像设备和照明变化有关，另一种与物体和背景中强度的基本特征有关。我们将第二种强度不均匀性称为强度复杂度。在本文中，我们专注于对具有强度复杂度的图像进行分割。我们的主要论点是量化复杂的强度并将它们转换为有用的特征以提高分割精度。引入了两个称为区域强度复杂性指数 (RIC-Index) 和因子 (RIC-Factor) 的新量来量化强度复杂性。然后将量化的强度复杂度合并到变分水平集框架中。所提出框架的总能量函数由以下三项组成：区域强度复杂度项、局部区域拟合能量项和基于边缘的能量项。第一项是通过利用图像的区域强度复杂性因子来定义的。局部区域拟合能量利用均值和方差来描述图像纹理信息。能量泛函的最后一项，也是从区域强度复杂度因子推导出来的，包含了重要的边缘信息。通过整合这三项，所提出的模型可以处理强度复杂度图像，特别是两种图像：一种具有复杂的物体强度，另一种具有复杂的背景强度。