Optical character recognition (OCR) is a fundamental problem that predates even the birth of computers. Over the decades, a huge bulk of research work has been published, with the majority focused on Roman-like and/or East Asian scripts. This survey paper, however, provides us with a comprehensive introduction to Urdu-like scripts.

In general, this survey presents rich background knowledge on the nature of Urdu-like scripts and their variants in read/write formats. By demonstrating those complexities with well-annotated figures and tables, the authors deliver a vivid lecture to people who are less familiar with the scripts widely spread throughout the Middle East.

One question I would like to ask is whether we have scientific evidence of the failure of applying existing techniques for well-studied languages, such as English, on Urdu-like scripts. What I usually observe in the literature are common techniques already used on other languages. For example, common methods include horizontal projection profiles for text line segmentation, structural features for classification, and other well-covered techniques.

On a side note, I believe the trend of OCR research is to aim for robustness on more natural and/or noisy inputs. Technical robustness includes less dependence on segmentation-based inputs and handcrafted features.