Digitization and Reconstruction of Archaeological Artifacts

We describe an ongoing research project on efficient methods for reconstruction of objects from large collections of irregular fragments, such as ancient pottery, collapsed murals, etc.. Our solution for flat objects uses multiscale matching and constrained dynamic programming, and we are now extending it to curved pottery fragments. We seek collaborative projects with museums and other potential users. 1. The fragment reassembly problem. Our project aims at developing efficient algorithms that can find adjoining pieces in a large collection of irregular fragments, by comparing their shapes. Besides archaeology, we expect these algorithms will be useful in art restoration, failure analysis, and other areas. For flat objects, the data typically consists of 10–10 irregular fragment outlines, each defined by 10 sample points. The goal is to find any pairs of curves that have long sections with similar shapes — as similar as expected from adjacent fragments of the same object: