A Study of Text Extraction Algorithms for Natural Scene Images

Extraction of text deals with detection and recognition of text from the given image. However, this extracted text may bring revolution in the field of computer vision if used in unrevealed domain when image clicked in natural environment images. One perspective is to use the extracted text in car driving application to get direction at the time of scorching sun or in foggy weather in winter. Our main focus is to enlighten the untouched area where computer vision can work better. To fulfil this, here we focus on algorithms, methods, approaches to pre-process, detect, and recognize text that could be applicable in natural scenes also. In this paper, an attempt has been made to discuss images with specular highlights, classify classical and deep learning-based approaches, and examine significant difficulties and methodologies to extract text. It also explores benchmark datasets and its evaluation processes for text extraction and specular reflection suppression.