Self-adaptive global and local double-layer optimized image generation model and generation method

The invention provides a self-adaptive global and local double-layer optimization image generation model GL-GAN and a self-adaptive global and local optimization method Ada-OP. A local double-layer optimization model is combined with a traditional global optimization model; according to the feature map output by a model discriminator, the quality measure of each area is obtained in the image; a low-quality region in a sample is optimized through accurate capture; local information in the feature map is taken as the basis of adaptive global and local double-layer optimization; a local double-layer optimization model is used for guiding optimization of a generator, so that the generator focuses on a global region and a local region of an image simultaneously on the basis of a maximum and minimum game of a generative adversarial network (GAN), coordination optimization is carried out on the whole and the local of the image, and a function of generating a high-quality image while high calculation efficiency is re