Research on Text Correction Technology Based on Secondary Transformation

In order to solve the complex situation, the OCR text image can be accurately corrected in multiple directions, through the analysis of the characteristics of OCR text, this paper proposes a stable and accurate OCR text correction technology based on secondary spectrum transformation and deep learning. According to the characteristics of the text, this technology first performs two spectral transformations on the text image, and then processes the processed image into a binary image to accurately extract the rotation angle of the text image and perform multi-directional correction on the text image. Experiments show that compared with traditional spatial domain Hough transform algorithms, this method has a faster, more stable and accurate correction ability within a certain text pixel size.