Recognizing text has been a challenging problem in computer vision applications. Extraction involves localization, detection, and recognition of the text from the given image. However, variation of text due to differences in size, style, orientation, alignment, low image contrast, and complex background make the problem of automatic text extraction extremely challenging.
The character recognition is done through stroke width transform which extracts the text region by improving complexity. It introduces new feature descriptors using skeletonization. The proposed methods are extensively evaluated on several benchmark datasets, namely ImageNet dataset for text extraction, Chars74k dataset to extract characters, and then recognized characters using K-NN classifier. The strength of the approach is efficient feature extraction which helps to achieve a recognition rate of above 90%.
https://ieeexplore.ieee.org/stamp/stamp.jsp?tp=&arnumber=7933250