Extraction of Handwriting Lines from Document Images Based on Concentration Distribution of the Local Area

Abstract

Paper documents have been widely used, and the digitization has been propelled for document management and efficiency. Additionally, since notes on them include users' information and interest, it is worth analyzing those information. However, it is difficult to perform character coding on both handwritten and machine-printed texts with an optical character reader. This paper proposes a method to extract the lines handwritten with writing tools from document images. The proposed method will facilitate the automation of the analysis and can be utilized for the extraction of important words and the preprocessing of handwriting information analysis. This method focuses on the concentration characteristics of writing tools. In the experiments using this method, handwriting and machine-printed lines were successfully extracted with 98% accuracy or above, and the false extraction of machine-printed areas was very little.