The digitization of images and documents has become an indispensable part of modern information construction. Aiming at the problem of image correction in the process of digitization of color art images, an image processing algorithm based on improved imaging model and deep neural network is proposed in this paper, which could effectively remove noise and increase image clarity. Firstly, an improved imaging model composed of global illumination, local illumination and reflectance is established, and the input RGB color image is converted into an HSV color image. Then extract the local features of the input image, and use these features to build a deep neural network to achieve noise removal. Simulation results show that compared with the traditional BP neural network and generalized regression neural network, the proposed algorithm has stronger image correction capability, higher peak signaltonoise ratio and contrast increment. Therefore, the proposed algorithm is feasible and advanced.