Tongue image segmentation method based on improved Unet network, electronic equipment and storage medium

The invention provides a tongue image segmentation method based on an improved Unet network. The tongue image segmentation method comprises the following steps: acquiring an image to be segmented; an image is input into the improved Unet network to obtain a coarse segmentation image, a convolution block in the improved Unet network comprises a multivariate branch module, a cascade layer comprises a space-coordination attention module, down-sampling is carried out by using a cavity convolution module, and up-sampling is carried out by using a Multi up-sample module; and inputting the roughly segmented image to a full-connection conditional random field module to obtain a precisely segmented binary image, and outputting the precisely segmented binary image. According to the method, a multi-element branch module is utilized to replace a convolution block in a traditional UNet network, so that the network can retain more detail information of a feature region in an image, an initial spatial position can be reserv