A Tool for Extracting Text from Scanned Documents and Convert it into Editable Format

Abstract

With the advent of Social media, now-a-days most of the data are stored in images. These data if processed correctly can provide large information. Hence there is a need to convert these data into information. A novel technique which can convert the data available in image into an editable format is proposed where the image can be acquired either by a camera, smart phone or directly from any source. The image is segmented into characters by using Connected Components (CC) and edge recombination using stroke width. This image is then converted to an editable format by using Optical Character Recognition (OCR) technology and Maximally Stable Extremal Region (MSER) for segmentation. The proposed system can also extract object from the images, which is done by using Artificial Neural Networks (ANN). The complete solution is developed using MATLAB and the output is stored in a variable. This can also be extracted to a word document if required where it can be edited. The performance of the system is measured by using two parameters namely precision rate and recall rate and has about 88% precision and 97% recall rate which is higher than most of the earlier proposed methods.