This report details the evaluation of current image matching implementations for the use in an image search engine, specifically for digital history. Due to the vastness of historical (digital) libraries this search engine must be able to search all (inter)national databases with equal performance. Current search engines use linguistic keywords to describe an image and search for others, introducing a language bias. This project focuses on image-to-image matching, bypassing language altogether.

This report only addresses image matching algorithms based purely on mathematics, no machine learning is addressed within this thesis. This report will cover the performance and usefulness of template matching, ORB feature extraction, SIFT feature extraction, SURF feature extraction, Brute-Force matching, and FLANN matching. Machine learning algorithms for the use in an image search engine are addressed by the other subgroup of this project.