Time variant data time super-resolution visualization method based on deep learning model

The invention relates to the technical field of data processing and modeling. The invention provides a time variant data time super-resolution visualization method based on a deep learning model, and the method comprises the steps: firstly extracting key voxels based on a gradient histogram, and reducing the data scale while maintaining the spatial features of original data; next, training a multi-scale variational auto-encoder to obtain an encoder with a feature extraction function and a decoder with a volume data generation function, so that the generation problem of a time variant data sequence can be converted into the generation problem of a hidden variable sequence. Therefore, the volume data can be processed and generated in the low-dimensional feature space which is more concise and can express potential information of the volume data. In the hidden space, two thoughts are provided to fit the time sequence relationship between the data hidden variables of each time step: in one method, the data is dir