The first part of this report proposes a linear camera model that generates images of the minimum mean squared error (MMSE) with respect to perspective projection images of a three-dimensional (3D) object. Error property of 2D images of the proposed camera model is compared to other camera models. In the second part of this report, we apply the proposed camera model to a iterative correction method for 3D shape reconstruction, and propose a more accurate shape reconstruction method within the framework of linearity. This method recursively composes a series of generalized inverse matrices (g-inverse) that starts with Moore-Penrose's g-inverse and recursively corrects reconstructed shape to achieve higher accuracy of 3D reconstruction. The authors have already proposed the same correction method for shape reconstruction by using weak perspective camera model, and have shown its validity I. The present report shows that the proposed method can reconstruct more accurate 3D shapes.
展开▼
