基于改进CV的图像分割

Image segmentation based on an improved CV model

为了提高传统CV分割算法对含纹理的图像分割效果,该文分析了图像纹理分量的统计特性,在图像TV平滑的基础上,设计了新的保边平滑函数,建立了保边平滑模型;运用传统CV分割算法对平滑分量进行分割;根据不同平滑分量分割区域不同,设计了基于区域置信度的分割平滑收敛函数,解决了分割曲线消失问题.实验表明,相对于传统CV分割算法,该算法对自然图像分割的效果较好,对非一致区域不敏感.

In order to improve CV model＇s performance on image with texture, this paper analyzes the statistical properties of image texture components, and proposes an improved segmentation algorithm. Specifically, this study has designed a new edge-preserving function based on the TV image smoothing. Also, the algorithm applies the CV model to segment the smooth component of an image. On the other hand, considering the fact that different smooth components have different segmentation areas, a new ＂Smooth convergence function＂ based on the region confidence is designed to overcome the disappearing of segmentation curve. Experimental results demonstrate that the proposed algorithm outperforms the traditional CV model in term of segmentation effects on natural images, but is insensitive to the inconsistent area.