文章基本信息

标题：A Survey on Text Recognition from Natural Scene Images and Videos
本地全文：下载
作者：Usha Yadav ; Nilmani Verma
期刊名称：International Journal of Soft Computing & Engineering
电子版ISSN：2231-2307
出版年度：2015
卷号：5
期号：1
页码：137-140
出版社：International Journal of Soft Computing & Engineering
摘要：Text recognition from any natural scenes images and videos is application of image processing technique. Basically text recognition is belongs to the pattern recognition which is part of image processing techniques. Now these days text recognition from natural scene images and videos is very difficult task.For make it easy four basic steps must be apply that approaches are (i) Text image pre processing (ii) character segmentation (iii) character recognition and (iv) Text recognition. In the state of art methods , character segmentation having two major approaches that is Segmentation –based approaches which segment the text into individual character before recognizing and segmentation-free approaches which recognizes character directly from whole text images without any segmentation. Character can also be recognized with two approach that is pattern matching methods in that particular method character are usually identified by a set of features and machine learning methods in which the methods are designed that are learn automatically from the image or after extracting feature. Various method has been applied earlier for extracting text from images and videos. These all methods are trying to provide better result .Various paper use printed text images for recognition their we never required any preprocessing for extracting text . Here is the name of some methods that are used for text recognition that are are Specific directed acyclic graph techniques, scalable feature learning algorithm, k nearest neighbour technique and back propagation algorithm. All those method which has been applying for text recognition , they all provide accuracy in result or we can say that the recognized text are nearly matched with the original one.
关键词：Neural based OCR ; Character segmentation;character recognition; Back propagation neural network model;Unsupervised learning