显著性驱动的局部相似拟合模型分割算法研究

Research on image segmentation based on saliency map and local likelihood fitting model

灰度不均匀和噪声图像的分割是计算机视觉中的难点。现有的活动轮廓模型尽管能够取得较好的分割效果,但仍然对噪声图像分割效果不理想,初始轮廓曲线的选取敏感,优化易陷入局部极小导致演化速度慢等问题。针对该问题,首先使用局部区域灰度的均值和方差拟合高斯分布,构建新的能量泛函,均值和方差随着能量的最小化过程而变化,从而增强了灰度不均匀和噪声图像的分割能力。此外,结合视觉显著性检测算法获取待分割目标的先验形状信息,并自适应地创建水平集函数,从而降低了初始轮廓位置敏感性及计算时间复杂度,实现全自动的图像分割。实验结果证明,提出的算法可以用于灰度不均匀和噪声图像分割,并取得了较好的分割性能,消除了算法对初始轮廓位置敏感性,减少了迭代次数。

Image segmentation with intensity inhomogeneity and noise image is still a challenge in computer vision.Although existing active contour models have been demonstrated to be an effective method for image segmentation, these models are sensitive to the initial contour of evolution curve and noise, which will tend to fall into local minimum with slow speed. In order to solve these problems, this paper presents a novel algorithm based on saliency map and local likelihood fitting model. Firstly, the local likelihood fitting model is constructed by describing the neighboring intensity with local Gaussian distributions, where the means and variances of local intensity in the energy functional will vary as minimum processing. Therefore, the proposed fitting model can improve the performance of segmentation for the images with intensity inhomogeneity and noise. Secondly, the prior shape knowledge of potential segmentation object extracted by visual saliency detection method is used to initialize level set function in order to reduce the influence of initial contour selection and the time-complexity. Furthermore, the proposed model can achieve fully automatic segmentation. Experimental results demonstrate that the proposed method can provide better segmentation for the images with intensity inhomogeneity and noise when comparing the existing active contour models. The proposed method also overcomes the problems that the existing models are sensitive to the selection of initial contour and have high time-complexity.