Toth : uma abordagem para extração de elementos textuais em imagens com variação de iluminação e linhas de texto inclinadas

Abstract

Optical character recognition software is designed to convert textual elements of documents into editable and searchable text. This task presents specific challenges when submitted to images captured by smartphone cameras. This work presents an approach to extract textual elements in images with variation of lighting and inclined text lines. The experiments demonstrated that the approach obtained a significant increase of effectiveness in relation to the baseline, at the moment that it also presented superior efficiency.