本文通过分析自然图像的边缘3个通道之间的关联性，提出“同一物体的边缘在3个颜色通道应处于相同位置”的色差先验约束，该约束在数学上近似为各通道的相对导数相等，基于此色差先验约束，建立了一种新的像差校正模型即图像解卷积模型，并给出了基于交替方向乘子法的模型求解算法。实验结果表明: 本文的像差校正技术可以提升图像的峰值信噪比10 dB以上，明显优于目前主流的BM3D和YUV算法，并且视觉提升效果明显，基本满足普通光学系统对像差的校正要求。

A priori constraint of the chromatic aberration of “the edges of the same object should be in the same position in the three color channels” is proposed by analyzing the correlation between the three channels of the natural image edge in this paper. The priori constraint is mathematically approximated as the relative derivative of each channel. Based on this chromatic aberration prior constraint, a new aberration correction model, namely the image deconvolution model, is established, and a model solving algorithm based on ADMM is given. The experimental results show that this aberration correction technique can improve the peak SNR of image by more than 10 dB, which is much better than the current mainstream algorithms such as BM3D and YUV. Moreover, the visual image performance is greatly enhanced, thus basically meets the common optical system correction requirements for aberrations.