一种基于犹豫中智集和水平集的图像分割方法

皮肤镜图像分析技术的第一步为图像分割,分割的结果会直接影响到后续的处理过程,针对具有背景噪声、模糊边缘和灰度不均的皮肤镜图像,提出了一种新的结合水平集的犹豫中智集图像分割方法;首先利用犹豫中智集理论将图像转换为犹豫中智集图像,其中犹豫中智图像由三类子集组成(T、I、F),利用犹豫中智集图像突出表达图像的目标信息和边缘信息;然后针对传统 DRLSE 水平集的不足进行改进,构造新的边缘停止函数,并增加灰度驱动能量项,最后通过改进的DRLSE 水平集对 ISIC(2018)皮肤镜图像进行分割测试;实验结果的交并比(Jaccard Index)值均大于 95%,且均方误差(MSE)、峰值信噪比(PSNR)和结构性相似指数(SSIM)均表现良好,表明方法能够准确、有效的分割具有模糊边缘和灰度不均的皮肤镜图像,对后续的皮肤镜图像的处理与诊断奠定了基础。

An Image Segmentation Algorithm Based on Hesitant Neutrosophic Set and Level Set

The first step of dermoscopic image analysis technology is image segmentation, and the result of segmentation will directly affect the subsequent processing. For dermoscopic images with background noise, blurred edges and uneven gray levels, a new hesitant neutrosophic set image segmentation method combined with level sets is proposed. In this method, the image is firstly transformed into the hesitant neutrosophic image by using the theory of hesitate neutrosophic set, in which the hesitate neutrosophic image is composed of three kinds of subsets (T, I, F), and the hesitant neutrosophic set image is used to highlight the target information and edge information of the image. Then, aiming at the shortcomings of the traditional DRLSE level set, a new edge stopping function isconstructed, and the gray driving energy term is added. Finally, the ISIC (2018) dermoscopic image is segmented and tested through the improved DRLSE level set. The Jaccard Index values of the experimental results are all greater than 95%, and the mean square error (MSE), peak signal-to-noise ratio (PSNR) and structural similarity index measure (SSIM) all perform well, indicating that the proposed method can accurately and effectively segment the dermoscopic images with fuzzy edges and uneven gray level. This research lays the foundation for the processing and diagnosis of subsequent dermoscopic images.