Offline Text-Independent Writer Identification Using Different Levels of Features
Dongli Wang · 2019
The effective extraction of feature information in handwriting identification has been the focus of researchers, especially the completion of a robust handwriting identification method is still an urgent technical problem. In this paper, we will combine two different levels of features: local directional chain-code feature (LDCF) and global improved texture feature (GITF). According to the advantages of each of these two features, it is applied to different matching processes. In the first stage, the LDCF is extracted and then roughly matched to obtain a handwriting image candidate sample set. The next stage is to refine the candidate sample set using the GITF. The experimental results evaluated on the database containing 203 writers of address images demonstrate the effectiveness of our method.