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像素值。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。由于局部指向性的计算在本质上是图像梯度的计算，因此采用几种不同的梯度算子来计算局部指向性。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，同时比较了不同算子在不同光照下对反向合成梯度算法的影响。

Improved inverse compositional algorithm and operators comparison

Traditional inverse compositional image alignment algorithm (ICIA) compares intensity values between a template image and an input image. However, it is easily affected by the variations of the lighting conditions, which leads to poor convergence or even divergence. Local orientation of an image tends to be less sensitive to lighting conditions than intensity values. Therefore an inverse compositional gradient (ICG) algorithm was proposed based on the local orientation. Experimental results demonstrate that this algorithm could overcome the influence of the lighting variations efficiently. In addition, the evaluation of the local orientation and image gradients is essentially equivalent. Hence several gradient operators were used to compute the local orientation, and the influences of the operators on the ICG algorithms under different lighting conditions were analyzed and compared.