Digital image distortion and proofreading technology in micro focal spot X-ray inspection

For the sake of the processing errors and constructional offsets of the lens of CCD camera, the digital image gained in micro-focal X-ray inspection has radial and tangential and other non-linear distortion. This paper presents an approach to proofreading the image distortion. First of all, a X-ray inspection image of a PCB with 19×19 BGA pads is taken as a calibration sample image, and is processed by the local adaptive thresholding algorithm. Secondly, corner points which are defined as the gray level change are the biggest points along horizontal and vertical direction within their adjacent fields respectively are collected by using Harris Corner algorithm with Gaussian smoothing coefficients. Thirdly, to reduce the computational complexity, some points are selected in a certain interval from the above corner point collection as the basis sample points of distortion correction model which using the least squares algorithm for two-dimensional N-order polynomial fitting, and the correlation fitting coefficients are calculated. Fourthly, the distorted sample image is corrected by using the distortion correction model. At last, to verify the universality of the correction model an image of X-ray detection is arbitrarily chosen and is corrected. Comparing the two images before and after correction processed it can be found that the presented proofreading approach has not only rather correction precision but also quite universality.