In the process of restoring the image of space optical camera, it is necessary to use the standard 24 color card to correct the image. In order to improve the accuracy of the mapping function in the process of color correction and prevent the occurrence of overfitting at the same time, a neural network correction method with comprehensive objective function is proposed. Using the idea of regularization as a reference, a constraint term with smooth performance is added to the objective function of the neural network to improve the generalization ability of the network map to data. The experimental results on the dataset of the real color card of the space camera sho-ore the correction.