Medical Image Character Recognition Based on Multi-scale Neural Convolutional Network
Xing Xu, Wei Wang, Quanli Liu · 2021 International Conference on Security, Pattern Analysis, and Cybernetics（SPAC) · 2021
Generally, medical images are usually characterized by complex background and low resolution. Also, text's location is sometimes random. Therefore, traditional Optical Character Recognition (OCR) technology cannot be well applied to recognize medical images which contain many factors. Aiming at this problem, a character recognition algorithm for medical images is proposed in this paper. The Efficient and Accurate Scene Text (EAST) scene detection algorithm is used in text detection and the Convolutional Recurrent Neural Network (CRNN) model is used in character recognition. The problem of low recognition rate that caused by the low resolution will be solved by adding a multi-scale convolution kernel to the model. The designed algorithm is tested by Medpix datasets, and the experimental results show that the proposed method has a higher precision and a higher recall than the traditional recognition methods.