The purpose of this research is the development of efficient strategies for multi-resolution object recognition. Multiresolution approaches are often used in computer vision, especially for tasks that are high computation intensive. We study the problem of recognizing two-dimensional objects described in multi-resolution by labeled trees and formulate the matching problem as an approximate tree matching problem. Sequential and parallel solutions to this problem are investigated.