Text Extraction and Recognition in Manga Comics Using Image Processing Techniques

Manga is a famous Japanese comic that spread the Japanese culture around the world. Translators, editors and designers around the world work on translating these mangas into many different languages to spread the Japanese culture and introduce them to other cultures. The time it takes to translate an existing Japanese manga is time consuming. Many approaches were suggested in text extraction but a few of them focused in mangas. In this work we propose an automatic method for the detection and extraction of characters within a manga comic page of all types; flat and non-flat. Several image processing techniques are utilized in our method. Text is extracted using the text extraction method followed by the OCR function, which is preceded by the frame content extraction method, which uses a blob extraction function. Our experimental results yielded 100% accuracy in extracting blobs from flat comic images, 90% from non-flat comics, and 90% accuracy in text extraction in both types. In many cases the proposed methodology is 90% faster than conventional methods.