Infrared weak and small target detection method and device based on diffusion model framework

The invention discloses an infrared weak and small target detection method and device based on a diffusion model framework, and the method comprises the steps: inputting a detection image into a trained dual-path coding and decoding neural network model, and predicting the noise in the detection image for multiple times through the model, and sequentially subtracting the plurality of predicted noises from the detection image to obtain a target image. The double-path coding and decoding neural network model comprises a plurality of double-path encoders and a resolver, the double-path encoders can extract semantic information and noise features in samples through two feature extraction paths during training, and the resolver can restrain high-frequency noise in the features. Therefore, the dual-path coding and decoding neural network model can fully learn more accurate semantic information and noise features of the sample, the target in the detection image is extracted through the trained dual-path coding and d