A Novel Database for Arabic Handwritten Recognition (NDAHR) System

Any proposed system for Arabic handwriting recognition systems requires a real and substantial database. In this paper, A Novel Database for Arabic Handwritten Recognition (NDAHR) is designed as it contains all sets of Arabic language. The original form of the proposed database consists of 6 pages, 24, 23, and 14 shapes in each page and 123 in overall of pages written by 100 writers divided into two groups. The first group consisted of faculty members and students in two universities in Iraq. The second group included employees in some government departments and some primary and secondary school students. After scanning all the forms by a high-quality scanner, we applied the cropping process using the Adobe Photoshop program to cut all images and classify them into different sets and then apply some methods of preprocessing to enhance the images. The collected database contains 61500 images divided into two groups: 80% training and 20% testing.