Fisher准则和正则化水平集方法分割噪声图像

随着活动轮廓模型的不断成熟和发展,模型的抗噪能力又成为重要的研究课题.为了精确地分割图像的同时去除图像的噪声,针对噪声图像用非负稳健函数作为边缘保持函数,从而保证图像在去噪的过程中边缘和纹理信息不被模糊.首先创造性地将分类器中的Fisher准则函数引入到图像分割中,从分类的角度对C-V模型给出了Fisher解释.把Fisher准则作为分割的标准来建立一个基于区域和边缘相结合的同时完成去噪和分割变分水平集分割模型.其次详细讨论了该模型的数值求解方法.最后实验验证了用Fisher值来衡量分割标准的理论的正确性和可靠性以及模型中正则项约束在去噪过程中的边缘保持功能.通过3组实验检验了提出的模型对噪声图像的去噪和分割比聚类算法、松弛迭代算法、Mean Shift算法有更好的效果.

Noise Image Segmentation Using Fisher Criterion and Regularization Level Set Method

With the active contour model gradually coming to maturity and development,how to improve noise immunity of the model has become an important research topic.Considering the image segmentation from the perspective of the classification,Fisher criterion function is introduced into image segmentation creatively,which uses Fisher criterion to guide the image segmentation and gives Fisher explanation about the C-V model.For the noise image,non-negative robust function is defined as edge-preserving potential function and ensures that the edges are not blurred.In order not to smooth out the edge and texture information in the process of removing noise,we introduce a robust function for regularization,regard Fisher criterion as the standard of image segmentation,and establish a variational level set model which is the combination of region-based model and edge-based model.The model can perform simultaneously denoising and segmentation.The numerical solution of the model is discussed in detail.Experiments confirm that the theory is correct and reliable,and the regularization term constraints are able to achieve good effect in the respect of noise removal and edge preservation during the process of image processing.The model for image denoising and segmentation is more efficient than that using clustering algorithm,relaxation iterative algorithm and mean shift algorithm with three sets of experiments.