A Novel Image Mosaic Method for Strip-Type Paper Fragments

摘要

Paper fragment mosaic is one of the applications of Image Mosaic, which plays an important role for Historical Document Re. Based on a series of modeling ideas, we propose a novel method to mosaic strip-type paper fragments In this paper. Firstly, the boundaries are sampled and binarized from paper fragments. Secondly, a matching ratio model is built to calculate the possibility that two fragments are spatial adjacent. Thirdly, the mosaic problem is transformed into a Traveling Salesman Problem (TSP) based on graph theory, and then solved via Linear Programming. In the experiments, a Chinese and an English paper are vertically cut into pieces, each of which is scanned as a picture file. The program coded based on our method can automatically mosaic all the fragments into the complete original paper picture, which proves the effectiveness of our method.