Low-illumination image enhancement algorithm model under attention mechanism

The invention discloses a low-illumination image enhancement algorithm model under an attention mechanism. Belongs to the technical field of deep learning. The method is characterized in that an attention mechanism and zero sample learning are introduced to improve image quality. The method comprises the following steps: firstly, designing a depth separable convolution enhancement factor extraction network to effectively estimate pixel-level light insufficiency of a low-illumination image; secondly, a loop image enhancement network is designed to be combined with an attention mechanism, and the low-light image is gradually enhanced according to the bearable model size; experimental results show that the algorithm can effectively improve the image quality, improve the definition of the image, avoid color distortion, achieve better effects in experiments for synthesizing the low-illumination image and the real low-illumination image, and well solve the problem of low-illumination image enhancement. 本发明公开了一种注意力机