Dynamic range compression method and system based on extremely lightweight Unet

The invention discloses a dynamic range compression method and system based on an extremely lightweight Unet, all data are processed in batches by using an adaptive gamma algorithm and Photoshop, details of a dark part of an image are better reserved, the dark part is properly brightened, overexposure of the bright part is avoided, and the processed image is natural and has no distortion or unreal color; the lightweight Unet network is adopted, compared with other deep learning algorithms, the model is light, the calculation amount is greatly saved, the operation efficiency is high, the flexibility is high, the effect is still prominent and stable in a scene outside a training set, most picture details are reserved, a brightness graph is used for training, then a weight graph is used as network output, and the method has the advantages of being high in robustness and high in robustness. A fitting mode is simple and easy to understand, contrast information of the image can be reserved, and excessive whitening