Deep Learning Model for Recognizing Text in Complex Images

Abstract

AbstractRecognition of text from complex background images is a challenging task due to the large variations in background, texture, font, and illumination conditions. Text can be detected easily from those images containing less background complexity. There are a few applications that detect the text from images. The accuracy obtained is low for the complex background images. To improve the accuracy, convolutional neural network is implemented. Basically, the proposed approach consists of three parts. At first, the text is detected from a complex background. Then the text is extracted from the image using Tesseract. Finally, all the detected words are stored in a text file. Then the text is converted into an audio file. The proposed system reads the text from the image with the aim to provide assistance for visually impaired persons.KeywordsImage processingDeep learningText recognitionComplex imagesVisually impaired