The Research On Digital Image Keystone Correction Algorithm And Application Of Video Real-time Correction

With the development of economy and science technology, computer vision system is widely used in transportation, medical, industry, military and other fields. Due to the limitation of filming site and imperfect device imaging system, the acquired image failed to meet the actual effect and came into being disproportional or deformation in the process of image acquisition. This phenomenon is known as the geometric distortion. Geometric distortion is mainly due to the displacement of pixels in the image, such as the objects in the image is out of shape, the distance ratio is imbalance and so on. The image distortion which appeared directly in the image, will produce great influence on the image analysis.This paper started from the research on keystone distortion of digital images, elaborated the principle of geometric distortion correction of digital images and the traditional calibration method, the circumstances of distortion is analyzed further and specific, and put forward two kinds of correction methods that applied to the images and videos. For the geometric distortion of image, the first step on the correction is histogram equalization to enhance the image; then using Sobel operator for edge detection on the horizontal or vertical direction; and by defining the angle and the Radon transform to extract the outline of distortion; finally, using the method of connection point for distortion correction. And correcting the distortion of video is based on the method of image correction, but due to the complexity of video is higher than that of image, the original distortion contour extraction is not always accurate. So a method is proposed to extract the edge of distortion through the human-computer interaction. This method can accurately and rapidly extract the edge of the distorted image and the corrected image is closer to the ideal image.A large number of experiments show that the restoration algorithm in this paper has high calculation efficiency, and the treatment effect is good. Furthermore the video correction algorithm has a good advantage of cross-platform transplantation.