Research Of Perspective Distortion Correction Method For Wide-angle Images

Nowadays, camera in mobile phone is more and more important, for the reason that social networking, selfie and photography has become the fashion. So the function of camere is the focus of competition among major mobile phone manufacturers. Most of these company has been adopted wide-angle lens which can get larger field of view. However when the wide-angle lens follows pinhole imaging model, the objects which are close to the camera in the imaging scene and located on the edge of the field would be seriously stretched due to perspective. Particularly when the object is a face or a person, the stretched distortion will be more significant due to eyes'sensitivity to face and person images. Although this distortion will not affect the clarity of the image nor distort line area of the image, it makes the characters or objects a serious stretched and greatly reduce the user acceptance of the photographs. Therefore, the development of correction algorithm for this distortion has great commercial value correction.For this kind of stretch distortion, firstly, this paper analyses the cause of perspective distortion and researches previous studies on the proposed correction method. Then we propose two methods, which are automatical content-adapting stretched distortion correction method and adaptive correction method based on image warping, respectively.The first one is a hybrid content-based image correction method. Our method automatically generates correction coefficient that adjusts to image content through face detection algorithm to keep lines in the background of images straight and maintain the foreground objects not to be stretched. Without any human intervention, our algorithm can automatically process distortion images. Simulation results show that the algorithm is convenient and practical, and the result of correction is visually well, too.In the method based on image warping, We establish a series of constraints which can correct the stretched distortion while prevent background straight line from bending. Then these constraints are expressed as energy functions. Least-squares method is applied to minimize energy functions in order to calculate the corrected coordinate value. MATLAB simulation results show that the presented method has good visual effect, which can be completely automated.