Image-text coding method based on depth decoupling framework

The invention discloses an image-text coding method based on a deep decoupling framework, and the method comprises the following steps: obtaining data to be coded, and separating image data and corresponding text data of each sample; inputting the image into a pre-trained image coding and decoding model, and coding the image to obtain an image vector; inputting the corresponding text data into a pre-trained text encoding and decoding model, and encoding the text data to obtain a text vector; combining the image vector and the text vector to obtain a vector with common features; through the common feature vector, an image specific feature vector and a text specific feature vector are decoupled; and finally, splicing the specific feature vector of the image, the specific feature vector of the text and the common feature vector to obtain a final image-text code. According to the invention, the image data and the text data corresponding to the image data can be coded at the same time, and high-quality image-text