郭晓冉, 崔少辉. 透视变换模型的Harris图像配准[J]. 微电子学与计算机, 2014, 31(4): 18-22,26.
引用本文: 郭晓冉, 崔少辉. 透视变换模型的Harris图像配准[J]. 微电子学与计算机, 2014, 31(4): 18-22,26.
GUO Xiao-ran, CUI Shao-hui. Image Registration Using Harris in Projective Transformation Model[J]. Microelectronics & Computer, 2014, 31(4): 18-22,26.
Citation: GUO Xiao-ran, CUI Shao-hui. Image Registration Using Harris in Projective Transformation Model[J]. Microelectronics & Computer, 2014, 31(4): 18-22,26.

透视变换模型的Harris图像配准

Image Registration Using Harris in Projective Transformation Model

  • 摘要: 针对图像末制导系统的图像配准问题,提出了一种透视变换模型下基于增强旋转不变性Harris算法的快速图像配准方法.首先,将Harris算法进行了旋转不变性的增强,利用增强后的Harris算法提取摄像系统中每帧图像的角点,采用基于归一化互相关(Normalized Cross Correlation,NCC)匹配算法进行角点的粗匹配;然后,利用随机抽样一致(Random Samples Consensus,RANSAC)算法对粗匹配的角点进行迭代筛选,剔除错误的匹配点,并将保留下来的精确匹配角点带入透视变换模型,计算出稳定的可反映图像平移、旋转和尺度变换信息的透视变换矩阵;最后,利用得到的变换矩阵对图像进行配准.实验表明,该方法对平移、旋转、尺度变化、视角变化以及光照变化和模糊变化均具有良好的不变性,可以快速高效地实现高精度的图像配准.

     

    Abstract: To resolve the problem of image registration in image terminal-guiding system,a rapid image registration approach based on Harris algorithm using projective transformation model is proposed.First,Harris algorithm is reinforced in rotational invariance,then corners of every frame in camera system are extracted using the improved Harris algorithm,and coarse matching is executed on corners using normalized cross correlation algorithm.Second,Random Samples Consensus algorithm is iterative used to filter the coarse matched corners,and removing the wrong matching points.The accurate matching points are put in projective transformation model to calculate the stable projective transformation.Finally,images registration is carry out by means of projective transformation matrix which can reflect the translation,rotation and scale change.Experimental results demonstrate that the proposed approach is invariant to translation,rotation,scale,viewpoints changes and robust to illumination changes and blur,and can be used to implement high precision image registration fast and efficiently under projective transformation model.

     

/

返回文章
返回