Research And Application Of Uneven Illumination Document Image Correction Method Based On Image Blocks

The rapid development of information technology and modern industry, and the rapid popularization of mobile devices such as smart phones and tablets, make more people use way of image taking pictures. Using computers, the Internet and other information processing technology to store and process images and text has become the mainstream. OCR technology is one of the most efficient methods of extracting texts from images, and is widely used. However, during the process of taking pictures, various subjective and objective factors such as poor lighting conditions, bad paper quality, etc., may lead to produce some noise, shadow and partial light in access to images. The interference can reduce image quality and directly affect the OCR recognition effect. At present the shortage of the OCR technology mainly for the low ability to adapt to low quality image, unable to extract full and right text messages.To solve this problem, this paper firstly introduced the basic theory of image processing field, then described and analyzed image features of uneven illumination text images, common used techniques in the field of image processing and some correction methods that widely used in domestic and international. After fully studied the regional brightness feature of uneven illumination text images, this paper proposed the correction method which based on image blocks.Firstly, this paper introduced the topic related to domestic and foreign research present situation and development tendency. Then expounded some image features commonly used in image processing field, basic theories of some key techniques and some popular algorithms including image enhancement, binarization and denoising. On the base of analyzed the common forms of text image illumination distortion deeply, the paper studied and expounded the effects of the illumination distortion on image features. The fourth part of this paper specifically described the study process and the implementation of the proposed correction method. First, the gray image blocked and calculated the brightness feature of each block. According to regional brightness feature of each block distinguish among normal image blocks and uneven illumination image blocks. For those normal image blocks, used the OTSU algorithm to binarize them to reduce the whole processing cost. And for other blocks, first did image enhancement on them, then calculated their regional brightness again to determine whether regional OTSU binarization algorithm or the improved White binarization algorithm would be used. The last step of the correction method was to do denoise process on the gotten binary image, to get better effect of the OCR. Test results showed that the proposed solution for uneven illumination text images can obtain satisfactory correction effect. It has advantages both in OCR recognition and correction time consuming than most of the existing correction methods. Addtionaly, the method has good robustness.