Grid constraint image splicing method and model construction method

The invention relates to a grid constraint image splicing method and a model construction method, and the method comprises the steps: building a deep learning-based unsupervised neural network image splicing model which is used for dynamically extracting image features from an input image, and enabling the trained model to be used for ancient wall painting image splicing. A powerful technical support is provided for application and popularization of a deep learning technology in cultural heritage digital protection in the future, and the problems of result artifacts and detail missing in an existing image splicing technology are solved.