Skin lesion image segmentation method and device, medium and equipment

Abstract

The invention discloses a skin lesion image segmentation method and device, a medium and equipment, and relates to the technical field of image recognition. The method comprises the following steps: after taking a skin lesion image, inputting the skin lesion image and preset noise into a pre-trained skin lesion image segmentation model; the model comprises a coding layer, a feature enhancement module, a decoding layer and an output layer. The encoding layer splices the image and the noise through a cascade encoder, extracts initial features, and fuses and updates the image and the noise; the feature enhancement module performs feature enhancement on the output of the last-stage encoder; the decoding layer combines the enhanced features and the jump connection information, and decodes to obtain a lesion subject image and a lesion edge image; and the output layer fuses the information to finally obtain a segmentation result of the skin lesion image.

Images