Text-zone detection and rectification in document images captured by smartphone

Abstract

Detection text-zone in the document images captured by smartphone is one of the main challenges in document image processing research domain. Unlike scanned document images, the document images captured by smartphones, especially under unconstrained condition and environment, have some additional challenges. The document frame may appear on various backgrounds, bent, rotated, light reflection, etc. To deal with these problems, we present a pre-processing method that allows to detect and rectify text-zone in the document images captured by a smartphone. The proposed method consists of three steps: binarization, skew angle detection/correction and text- zone detection. The experimental results based on the recognition rates given by two end- To-end OCR systems have shown a significant improvement when applying the proposed pre-processing methods before applying any OCR system.