Binarization Method For Non-uniformly Illuminated Document Images

The purpose of the early digital image processing was to improve image quality，butnow how to get the information that people needed is becoming the main purpose ofimage processing. In the automatic identification process of the text image characters, themost important step is binarization.The binarization results will affect the correct rate ofthe automatic identification.In the acquisition process of the text image, the non-uniform distribution of thebrightness will affect the quality of images. The binarization method is different, thetreatment effect of the non-uniform illumination is different. Many methods lead theresults loss of part of text or cover up by the black area causing more errors in therecognition result and the low recognition rate. How can a non-uniform illuminationimage be successful processed by binarization is worthy of study.In this paper, the binarization method based on curvelet transforms is to remove theinterference caused by non-uniform illumination. First curvelet transform to the images,and then get the curvelet coefficients. According to the characteristics of the curveletcoefficients, nonlinear enhance the curvelet coefficients in order to optimize the histogramof the image. Inverse curvelet transform to the optimized coefficients, and then get thetime-domain image. Finally, Otsu transform to the time-domain image to get binary image.The experiment results show that, compared with other classical binarization methods inimage quality, OCR recognition rate and computing time, the proposed method is moreeffective in processing non-uniformly illuminated document images.