高级检索

基于形状保持的CV变分水平集矩形目标分割

Rectangle Object Segmentation Based on Shape Preserving and CV Variational Level Set

  • 摘要: 目标在被局部遮挡、与背景灰度信息相似以及纹理比较明显等情况下,传统CV模型无法进行准确分割.为此,将模型中活动轮廓曲线的水平集函数用先验形状的水平集函数来代替,使得曲线在演化过程中始终保持某一类特定形状,从而实现了目标分割过程中的形状保持.根据形状保持的CV变分水平集分割模型建立适用于矩形目标分割的能量函数模型,推导出一组Euler-Lagrange常微分方程;通过水平集函数的不断迭代演化最终实现了矩形目标的分割;最后演化得到的水平集函数是对矩形目标的定量描述.3组实验结果证明,该模型能够解决复杂情况下的矩形目标分割问题,且具有计算量小、鲁棒性强的优点.

     

    Abstract: CV model is difficult to precisely segment the object which is partially occluded or is similar in gray value with the background or has obvious textures. In this paper, we add shape restraint equations of prior shape to the level set function, which keeps the curve to be a specific class shape during the whole evolvement and realizes shape preserving in object segmentation. Using the proposed model, we built the energy function for rectangle object, deduce a group of corresponding Euler-Lagrange ordinary differential functions, and evolve the level set function. By evolution, rectangle object can be properly segmented, and the final level set function is convicted just the quantitative description of the rectangle object. At the end of the paper, three groups of experimental results validate that the proposed model can correctly segment the rectangle object from complex backgrounds with lessened calculation and strong robustness.

     

/

返回文章
返回