Research On Low-Altitude Aerial Image Correction And Enhancement

Low-altitude aerial image is playing an important role in the living and production, such as road surveillance, disaster relief and emergency, landscape expedition and so on. The characteristics of wide-angle lens are big view, wide field of vision, long depth of field etc. When wide-angle lens is used in low altitude aerial photography, it can reduce the frequency and time of aerial photography. However, due to the structural parameters and the aircraft movement during shooting, there are wide-angle motion blur and other defects in the captured images. These defects reduce the accuracy of the low-altitude aerial images, which limits its application. In addition, as a single image contains very little information, we need to splice multiple images into an image. This image contains a large amount of information, so that we can observe and get information.In order to solve the issues mentioned above, the wide-angle distortion correction, motion blur and image stitching are studied and improved in this thesis. The main work is as follows:1. The study the structure and corresponding mathematical model of wide angle distortion has been made. On the basis of the existing basic method of the wide angle distortion correcting, the evaluating and analysis of the performance has also been made. This thesis proposes a segmentation correction method based on pitch search within a single image. This method chooses grid template, and uses the relationship between actual size of the grid spacing and distortion size of the grid spacing to determine the distortion center. It is easy and convenient. According to the different sizes of the distortion, we propose a correction method of segmentation.2. According to the characteristic that the spectrum of the motion blur image perpendicular to the direction of motion, the linear detection of image spectrum has been made. In this thesis, Radon linear detection is used herein to obtain the spectrum direction of motion image. The vertical direction is direction of motion. At the same time, the fuzzy length is obtained by the relationship of the spectrum, thus, we can restore the blur image according to the moving direction and the fuzzy length.3. According to the basic principles and steps of image stitching, we complete the image stitching and block output, we also improve the fast flowing water output method of image stitching. Firstly, the input and output methods of image stitching have been improved. The water models have been used in image stitching process to realize input and output synchronously and change traditional overall output and input mode into single input and multi-output mode. Secondly, we improve the size of the reference image in image stitching process, which makes the reference image stay the same size with the adjacent two images in the stitching process, so that the scale of the pieces of the stitching image is consistent. Finally, we improve the method of the image fusion. The seam is applied to image fusion to improve the quality of fusion and eliminate the seam. Multi-output in this thesis means the block output of stitching image, which makes it easy to store images and post-process the images.