Embodiments of the present disclosure disclose an
image processing method and device,
electronic equipment and storage medium, wherein the method comprises: performing the following steps by a
Transformer model: combining each window image of a to-be-processed image and window
feature data to be learned corresponding to each window image to obtain each combination data; performing self-attention transformation of pixel dimensions on each combination data to obtain intra-window
information exchange images and learned window
feature data; determining influence weights between the intra-window
information exchange images according to the learned window
feature data; and performing self-attention transformation of window dimensions on the intra-window
information exchange images according to the influence weights to obtain inter-window
information fusion images. The
image processing can be performed based on a lightweight
Transformer model, the calculation amount can be reduced on the basis of ensuring the
processing effect, efficient and rapid visual
image processing is realized, and the deployment of different platforms is facilitated.