Processing Images of Handwritten Documents Using Computer Vision Methods

Automation of handwritten document analysis is solved using image processing technologies. Handwritten texts are a complex object for information extraction due to the variety of writing styles, different levels of clarity and the presence of noise. Noise can manifest itself in the form of defects, blurriness or distortion. A cell on paper can also interfere with text recognition by creating visual noise, distracting the attention of recognition algorithms and worsening the overall accuracy of information extraction. The article considers several approaches to removing a cell on a sheet of paper using the OpenCV library of the Python: finding Canny boundaries, the Hough transform algorithm, the method of filtering an image by color characteristics and a combination of methods.