Abstract
A 3D model representing an object in three dimensions is employed in a variety of domains such as object analysis, simulation, and medical model. Currently, a 3D model reconstruction from a 2D scene image becomes an alternative way for interactive object modeling. In this field of work, 3-Sweep and Loft are examples of successful methods. However, these methods require user guideline or 3D model reconstruction, making it inconvenient for modeling each time. This paper proposes an automatic 3D model reconstruction consisting of two main steps: (i) closed contour construction is to extract suitable outline using local binarization. This phase can reduce interference of noise and texture to find the more accurate outline, and (ii) automatic guideline construction is to construct a shape guideline from the outline. The experimental results show that the closed contour construction can extract the shape object nearly similar to the reference one, especially when compared with the results of state-of-the-art methods, namely Sauvola, Bernsen, Singh, mean, and median thresholding methods. Furthermore, the 3D model reconstructed by the proposed method is empirically similar to the one reconstructed by user guideline.
Ref: http://ncit2016.wu.ac.th/proceedings/proceedings_ncit2016.pdf