Research On Related Technology Of Correction Of Distortion Of The Large View Camera

With the development of machine vision technology, computer vision has played an important role in people's production and life. The optical system can take more people and things when the viewing angle is relatively large, but it also have some flaws in imaging, which is we call distortion. The correction of distortion has great significance in the field of computer vision. Distortion is divided into two parts, one is the distortion which is caused by the optical system, and we call it optical distortion(geometric distortion). The other is produced by perspective of pinhole camera, for example perspective distortion and stretching distortion. The main purpose of this paper is to correct these distortions, especially the stretch distortion which is the most difficult. Since the distortion of the image is distorted by the mixture of the two distortions, so we correct the two aberrations respectively.For the correction of optical distortion, this paper describes some preprocessing methods such as calibration and corner detection. We also give the processes of optical distortion correction algorithm and results. After the optical distortion corrected, perspective distortion will revealed when human face is very close to the camera. We present a method that the axis of the points can be shrunk to the center point by the law of Gaussian Function and we have achieved good results.After we correct the optical distortion, the stretching distortion will appear when human face or other objects are on the edge of the view. Firstly, this article uses the method of controlling parameters and multi-plane projection, and we also achieved certain results. We propose new ideas of establishment of three-dimensional model, establishing and restoring the three-dimensional scene. The human face is approximately modeled as a three-dimensional piecewise Gaussian Function model, and the rest of the scene can be regarded as a plane. So we obtain the whole model. Then we move the camera so that the optical axis aligns to the human face. We achieve the final result by stitching and synthesizing. After correcting we also compare the effects of the different parameters of Gaussian model, different sizes of Gaussian Function, and different face models. While our paper also present the indicators of the objective assessment to evaluate the effect, and prove the correctness of our approach. Finally, this article also presents two stitching methods, introduces several matting methods, and shows the effect of synthetic splicing.