A Flourished Approach for Recognizing Text in Digital and Natural Frames

Abstract

: Acquiring tenable text detection and recognition outcomes for natural scene images as well as for digital frames is very challenging emulating task. This research approaches a method of text identification for the English language which has advanced significantly, there are particular difficulties when applying these methods to languages such as Bengali because of variations in script, morphology. Text identification and recognition is accomplished on multifarious distinct steps. Firstly, a photo is taken with the help of a device and then, Connected Component Analysis (CCA) and Conditional Random Field (CRF) model are introduced for localization of text components. Secondly, a merged model (region-based Convolutional Neural Network (Mask-R-CNN) and Feature Pyramid Network (FPN)) are used to detect and classify text from images into computerized form. Further, we introduce a combined method of Convolutional Recurrent Neural Network (CRNN), Connectionist Temporal Classification (CTC) with K-Nearest Neighbors (KNN) Algorithm for extracting text from images/ frames. As the goal of this research is to detect and recognize the text using a machine learning-based model a new Fast Iterative Nearest Neighbor (Fast INN) algorithm is now proposed based on patterns and shapes of text components. Our research focuses on a bilingual issue (Bengali and English) as well as it producing satisfactory image experimental outcome with better accuracy and it gives around 98% accuracy for our proposed text recognition methods which is better than the previous studies.