Estimation of 3-D parametric models from shading image using genetic algorithms

Abstract

In this paper, a method for estimating parameters of a 3-D shape f r om a 2-D shading image using a genetic algorithms (GAs) is proposed. The shape of the object is represented by a superquadrics model, and then the model parameters are c oded for being applied t o GAs. The coded string is evaluated a c cording to the similarity of the shading image calculated f r om the 3-D m o del shape r epresented by the parameters to the given 2-D shading image. By applying the GAs to the optimization of the evaluation value, the string having the minimum dierence c an be found. The parameters are estimated f r om some shading images of various 3-D shapes by using the proposed method, and the results are p r esented.

3D-spline reconstruction using shape from shading: Spline from shading

“…The methods described in [3,26,27,29], respectively, in [3,23,24,31,32], more or less successfully grapple with Pb1, respectively, with Pb3. On the other hand, Pb4 was not addressed in these works and only [31] dealt with Pb5, to some extent.…”

“…In [31], a superquadrics is used, whose 10 parameters are estimated via genetic algorithms, but no convincing result is shown. On the other hand, the results reported in [32] are of good quality, even if the method is intrinsically dedicated to face reconstruction.…”

A Global Solution to the SFS Problem Using B-spline Surface and Simulated Annealing

“…This problem has been partially solved in [12], where the same model is implemented, but using a non-regular grid and a multiresolution process. In [11], a superquadric shape with ten parameters is used, but no convincing result is shown. On the other hand, the results reported in [1] are of good quality, even if the method is intrinsically dedicated to face reconstruction.…”

“…In [3], a multiresolution approach coupling SA and deterministic optimization has been successfully designed, but is very slow, due to the huge number of unknowns. In [11], a genetic algorithm is used, but no experiment on real images is proposed.…”

Genetic algorithms for 3d reconstruction with supershapes

“…Among those methods, superquadric model [1] have been very well studied [2][3][4][5][6][7][8][9][10][11][12][13][14][15], when methods based on the recent supershape model [16] have been barely investigating [17,18]. Although an analogy between superquadric and supershape models can be made, and that optimization methods to recover the parameters of the best superquadric fitting are usually based on the LevenbergMarquardt (LM) theory [19], LM cannot be applied to retrieve all the parameters for supershape fitting.…”