In this paper we present a new variant of ICP (iterative closest point) algorithm based on local feature correlation. Our approach combines global and local feature information to find better correspondence sets even for the case of large displacements between model and image data. For such cases, the model points are aligned to image points (rotation plus translation in 2D) before the feature extraction process. This avoids the need of a normal pre-alignment step. Our approach was tested on synthetical and real-world data to compare the convergence behavior and performance against the original ICP algorithm with a classical pre-alignment approach.