Segmentation on medical noise images is a very challenging research task. This paper proposes an image segmentation model based on decomposition and active contour for simultaneous image de-noising and segmentation. The model is a variational functional integrating the image decomposition model in G space and the active contour model combining the boundary and regional information. The model solution can be decomposed into solving two functional extremes-image decomposition part and segmentation part to avoid the difficulty of searching for a direction solution, the image decomposition part solution is a functional extreme in G space and can be solved by threshold shrinkage in the second generation curvelet. The segmentation part solution is a variational level set functional extreme and its corresponding Euler equations is the nonlinear partial differential equation which can be solved by gradient descent flow. Experimental results show that the suggested model in this paper not only de-noises effectively but also has better segmentation performance than the Chan- Vese model, snake model, variational level-set model and the ASM under the same experimental conditions. Our method improves image quality and separates target part well.%医学噪声图像的分割是一件非常困难的事情,为了同时进行噪声去除和图像分割,提出一种基于分解的图像活动轮廓分割模型.该模型是G空间图像分解模型和边缘、区域相结合的活动轮廓模型集成的一个变分泛函,由于模型直接求解困难,把它分裂成2个泛函极值——图像分解部分和图像分割部分.其中,图像分解部分是在G空间的泛函极值,用第二代曲波变换域的阈值收缩求解；分割部分是变分水平集泛函极值,其Euler方程为非线性偏微分方程,可用梯度下降流求解.实验结果表明,文中模型不但可对噪声图像去噪,而且在相同的实验条件下分割效果优于Chan-Vese模型、Snake模型、Level-set模型和ASM；不仅提高了图像的质量,还能较好地分割出目标部分.
展开▼
