高级检索

自由差分运算与直交型Snake模型

Free Difference Operation and Orthogonal Snake Model

  • 摘要: 自由差分运算能够突破传统微分算子的模板运算模式,把对图像这种二维信号的微分计算转化成一维的方式进行,使计算方向可以根据实际问题灵活选择 直交型Snake模型基于任意曲线可由多边形进行逼近的思想,借助自由差分运算,把本应沿曲线法线方向计算的灰度梯度转化成在多边形各边的直交线上分析灰度的相对变化或绝对变化 在曲线的演化过程中,该模型以目标曲线的特征为指导,以基本多边形为依托,通过在直交线上进行差分操作,从而把轮廓变化的信息和图像灰度变化的信息联系在一起,使得所提取的轮廓曲线的特征不断地逼近目标曲线的特征 与已有的Snake模型相比,直交型Snake模型通过强调高层视觉和底层信息之间的相互联系在曲线演化过程中的作用,使得信息的处理流程处于闭环状态 在对畸形物体的轮廓提取实验中,获得了比较理想的结果.

     

    Abstract: Free difference operation breaks the template mode of classical differential operators. It turns the 2-D difference computation of image into 1-D difference operation. Therefore, the computing direction of difference operation can be decided by the actual problem flexibly. Based on the idea that any curve can be approached by a certain polygon, orthogonal Snake model turns the computation of gray gradient along the normal direction of a curve into the analysis of the relative or absolute variation of gray along the orthogonal line of each side of polygon by free difference operation. During each stage of curve development, this model can associate the contour variation with the gray variation by making difference operation along the orthogonal line, directed by the characteristics of the object curve and relied on the basic polygon. So the characteristics of the extracted contour approximate the characteristics of the object contour step by step. Compared with the conventional Snake model, the orthogonal Snake model stresses the affiliation between high-level information and low-level mechanisms, and its information processing is in closed state. Experiments on the contour extraction of malformed objects generate ideal result.

     

/

返回文章
返回