正则化方法在线性SFS问题中的应用

线性ShapefromShading问题(LSFS)是一类特殊的ShapefromShading问题,此时反射图是表面梯度分量的线性组合的函数。文章在待恢复表面平滑和已知光照方向等假定下,把线性SFS问题正则化,并利用Kaczmarz算法求解得到线性方程组。此种方法能处理区域形状不规则和边界条件不完备的情况。文章用Kaczmarz算法给出了一种从不可积向量场求得最接近的可积向量场的方法,该方法能处理区域形状不规则的情况。

Application of Regularization in Linear Shape from Shading

Linear shape from shading(LSFS)is a special kind of shape from shading,where the reflectance map is a function of the linear combination of surface gradient components.Under the assumptions of smoothness and known illumination direction,LSFS is regularized,then the resulting linear system is solved using Kaczmarz's method.This method can deal with irregular regions and incomplete boundary conditions.Based on Kaczmarz's method,an approach to compute a least-squares fit of integrable slopes to nonintegrable slopes is given,which can deal with irregular regions.