Meta-Heuristics for Solving Curve Fitting Problems in Optical-Diffraction Based Image Depth Reconstruction

Abstract

This paper investigates use meta-heuristic to solve curve fitting problems in Optical-Diffraction Based Image Depth Reconstruction. We aim to accurately establish a relationship curve between object distance and diffraction blur kernel through discrete data points. A mathematical model is first developed to formulate the concerned problems. Then, four meta-heuristic algorithms, particle swarm optimization, artificial bee colony, genetic algorithm and differential evolution, are employed and improved according to the problem’s feature. Finally, experiments are executed on a data set. A comparison of the similarity of the least squares method verifies the feasibility of meta-heuristics for solving curve fitting problems. Among the four meta-heuristics, differential evolution algorithm has the best competitiveness.