A single-image-based three-dimensional (3D) reconstruction method of an axisymmetric forge piece is proposed. The imaged circular points are solved from two paralell cross sections extracted from the image. Combining with the epipolar constraint, the 3-parameter constraint equations of the imaged absolute conic are established and the intrinsic parameters are solved. The imaged meridian and latitude of the surface of the axisymmetric forge piece are extracted based on the planner projection characteristics between two paralell cross sections. Then 3D shape of the axisymmetric forge piece is reconstructed based on the corrected imaged meridian. The visible region of the cross section is computed based on Laguerre formula, and an equal-arc sampling strategy is proposed based on the extracted imaged meridian, solving the inconsistent resolution problem of conventional texture extraction algorithm. Experimental results indicate that the large axisymmetric forge piece can be reconstructed with its own geometrical constraint from sinlge image without any auxiliary calibration equipment and the reconstructed shape and effective texture can be obtained.