基于支持向量回归的灰度图像三维形状重构

为实现未知光源的灰度图像高效重构形状,提出一种基于支持向量回归机和粒子群优化算法相结合的灰度图像重构三维表面形状的方法。通过研究和分析灰度重构形状(SFS)问题,基于支持向量回归机理论,构建了物体表面形状与其灰度图像间的非线性映射模型。对未知光照方向的实际图像进行光源方向估计,生成对应光照方向的训练样本以提高任意光照方向下的图像的形状恢复精度。为克服支持向量回归机中各参数选取无依据的不足,引入粒子群优化算法主动对各参数进行飞行寻优,使得回归模型为最优,以提高形状重构精度。最后,通过实例分析验证了所提方法的可行性及有效性。

Shape from Shading Based on Support Vector Regression Algorithm

In order to realize the shape from shading (SFS) under unknown light source parameters of single grayscale image, an efficient SFS method was proposed based on support vector regression(SVR) and particle swarm optimization(PSO) algorithm. Based on the SVR theory, a nonlinear mapping model was constructed between the grayscale image and its 3-D surface by researching into the SFS problem. The light source of the tested actual image should be estimated to generate the training samples corresponding to the light direction. It was important to select the proper SVR parameters for improvement on 3-D reconstruction precision, and the PSO algorithm was introduced. Finally, the case study had verified the feasibility and effectiveness of the proposed SFS method.