A New Framework in Interpreting Two-Dimensional Line Drawing

Abstract

This paper proposes a new framework to interpret a two-dimensional line drawing that represents a threedimensional object. The framework synthesizes three algorithms namely thinning algorithms, chain code algorithm based on Freeman Chain Code (FCC), and a new corner detection algorithms that accept chain code series as its input. Linear system of two image regularities namely spatial structure and skewed symmetry is identified and created. Total least square method namely Bisection and Secant method has been used to solve the linear system. The solutions obtained are the depth-value of each visible junction. Matlab functions are created to represent and solve the linear system, and to visualize the three-dimensional object derived. Three stages identified in the new framework are pre-processing, two-dimensional feature extraction and deriving depth values of each junction. Experimental result shows that the new framework sucessfully interpret a cube and produce threedimensional data. The comparison to the previous works on three-dimensional reconstruction is also discussed.