Low-light image enhancement method of image hierarchical structure network based on stream learning

The invention relates to the technical field of low-light images, in particular to a low-light image enhancement method of an image hierarchical structure network based on stream learning. The method comprises the following steps: forming an FI HN structure by an image hierarchical structure network and a reversible network based on normalized stream learning; and training the network by using a negative log-likelihood loss function. The invention provides a low-light image enhancement method of an image hierarchical structure network based on stream learning. The FI HN is composed of the image hierarchical structure network and a reversible network based on a normalized stream. The image hierarchical structure network extracts image information from three dimensions of local, regional and global so as to obtain image information as much as possible; under the constraint of a negative log-likelihood loss function, the reversible network based on the normalized stream can fully learn complex distribution of no