Research And Implementation Of Image Barrel Distortion Correction

In order to get bigger angles of view while taking pictures, it's necessary to add wide-angles lens, but inevitably this will introduce barrel distortion of pictures, i.e. convex deformation which is the same as fish eyes will be introduced to the snapshot, and this will affect the image processing. To restore the initial pictures, various techniques are needed to correct the barrel distortion.Currently the main approach to process barrel distortion is one is to correct the pictures with barrel distortion by image processing techniques. This method includes 2 ways, one is using planar template and the other one uses particular parameters of camera or the specific characteristics of image itself.The first using planar template method predefines a special template in the first instance, and then gets its standard image and distortion image; by using the pre-set special relevance (such as template certain points of corresponding coordinates) before and after the distortion, the mapping relation between two image pixels is finally obtained for picture correction. The method based on the particular parameters of camera is equivalent to sphere model method, which needs to know the focus of lens in advance, and then correct the distortion by using a certain arithmetic; The method based on the characteristics of image itself aims at the images taken are all belong to certain types, such as text image. This paper brings forward a method based on text characteristics:firstly expansion of lines of image text, lubricate and gets the middle text lines, then obtains image distortion correction parameter by simulating circular template, and eventually correct the images as a whole.The paper is based on the text image processing in project "Portable Reader", and explains the effectiveness of the proposed method by comparing OCR recognition rate after text image correction processed by many methods. Experimental results show that we can put the template aside to correct the distorted text image automatically, which don't reduce recognition rate, and also improve the adaptability of barrel correction.