基于视觉显著性的变分水平集图像分割仿真  

Variational Level Set Image Segmentation Simulation Based on Visual Saliency

作　　者：刘越[1] 薛佳楣[1] 韦韫韬[1] 李美珊[1] LIU Yue;XUE Jia-mei;WEI Yun-tao;LI Mei-shan(College of Information Science&Electronic Technology,Jiamusi University,Jiamusi Heilongjiang 154007,China)

机构地区：[1]佳木斯大学信息电子技术学院,黑龙江佳木斯154007

出　　处：《计算机仿真》2021年第6期391-395,共5页Computer Simulation

基　　金：佳木斯大学科研项目(JMSUJCMS2016-009);黑龙江省教育厅基本科研业务费基础研究项目(2016-KYYWF-0560)。

摘　　要：根据传统的水平集方法,在图像分割时,很难准确分割灰度不均衡的图像。提出了一种基于视觉显著性的变分水平集图像分割方法。通过Lab颜色空间建立视觉显著性数据模型,加强图像对比度,提取视觉显著性区域,随后对此区域图像做去燥处理,防止图像内细节信息的丢失。再利用无需初始化的变分水平集,结合能量函数完成内部能量的约束,得到曲线演化的偏微分方程,通过最小值能量函数,完成图像分割。仿真结果表明:所提方法解决了图像过度分割的现象,能够更快速的分割图像,具有稳定性、高效性和优质的鲁棒性。According to the traditional level set method, it is difficult to segment the image with uneven gray level accurately. Therefore, a method of variational level set image segmentation based on visual saliency was proposed. The visual saliency data model was built by Lab color space, and the image contrast was strengthened. Then, the visual saliency area was extracted. After that, the image of this area was denoised to prevent the loss of detailed information in an image. Moreover, the variational level set without initialization was combined with the energy function to complete the internal energy constraint, so that the partial differential equation of curve evolution was obtained. Finally, the image segmentation was completed by minimum energy function. Simulation results show that the proposed method improves and perfects the phenomenon of image over-segmentation, which can segment the image more quickly. In addition, this method has high stability, good efficiency, and high-quality robustness.

关 键 词：视觉显著性 变分水平集 图像分割 能量函数 偏微分方程 

分 类 号：TP391[自动化与计算机技术—计算机应用技术]