基于局部指向性的反向合成图像对齐算法

传统的反向合成图像对齐算法比较的是模板图像与输入图像之间的象素值。该方法容易受到图像中光照变化的影响,从而导致收敛性变差甚至发散。根据局部指向性对光照变化不敏感的特性,提出了一种新的反向合成图像对齐算法——反向合成梯度算法。介绍了局部指向性的定义及表达式,对反向合成梯度算法进行了详细的推导,总结了该算法的计算步骤。通过实验,验证了反向合成梯度算法能够有效克服图像中光照变化的影响。

A Novel Algorithm for Inverse Compositional Image Alignment Based on Local Orientation

Traditional inverse compositional image alignment(ICIA) algorithm compares intensity values between a template image and an input image.However,it is easily affected by variations of the lighting conditions,which would lead to poor convergence or even divergence.The local orientation is not sensitive to the lighting.Therefore we proposed the inverse compositional gradient(ICG) algorithm,which is a novel ICIA algorithm based on the local orientation.We introduced the definition and representation of the local orientation,derived the ICG algorithm and concluded its iteration steps.We used Photoshop to render the image in Fig.1 to imitate different lighting conditions.We considered three cases:(1) there is no other light but the ambient light with variable brightness;(2) besides the ambient light mentioned above,there is a spotlight with fixed position and brightness;(3) there is a spotlight with variable position and brightness and the brightness of the ambient light keeps constant.Experimental results given in Fig.2 show clearly that the ICG algorithm gives more accurate and reliable matching results under different lighting conditions.