Comparative Performance Analysis of Enhancement Methods Applied to Arabic Manuscripts

摘要

The ancient Arabic manuscripts are considered to be more complex regarding enhancement compared to others written in other languages. Complexity comes from character degradation, stains, low-quality images, curves of the text, character overlapping, etc. To facilitate the restoration, a set of well-known binarization techniques designed for historical document images is presented in this paper. Existing binarization techniques focus on either finding an appropriate global threshold for the whole image or adapting a local threshold for each area to achieve better enhancement quality. This improvement aims to remove noises, strains, uneven illumination, etc. The goal of our work is to assess these methods when applied to Arabic manuscripts in terms of readability, elimination of original spots and production of unwanted noise. Results show that no techniques work well for all types of manuscripts, but some techniques work better than others for particular types. Experimental results also indicate that Nick and Wolf techniques performed the best in terms of readability in most of the processed manuscripts.