An Efficient Hybrid Search Algorithm for Robust and Accurate Image Alignment under Non-uniform Illumination Variations

摘要

Image alignment is the most crucial problem in industrial visual inspection. Traditional intensity-matching based methods, including the normalized correlation method, are not robust against non-uniform illumination variations. In this paper, we present a generalized intensity-based matching approach to accomplish accurate and robust image alignment under high-level noises and large non-uniform illumination variations. This generalization is an extension of our previous FLASH (Fast Localization with Advanced Search Hierarchy) algorithm for image alignment. This is accomplished by matching the Laplacian-of-Gaussian filtered logarithmic (LoG-log) intensity function. The proposed hybrid search algorithm consists of two stages; i.e. the hierarchical nearest-neighbor (HNN) search and the optical-flow based energy minimization alignment. The HNN search finds the nearest-neighbor feature vector among a set of training data that is obtained by simulating geometrical transformation on the reference image with the geometric parameters uniformly sampled from a specified search space. The nearest-neighbor solutions give candidate pose estimates for the input image. Then the optical-flow based energy minimization method iteratively refines the pose estimation. Note that there are two modes for the optical-flow energy minimization alignment, i.e. the raw intensity matching mode and the LOG-log intensity matching mode. Some experimental results are given to demonstrate the efficiency as well as the robustness of the algorithm.