Dermatoscope image segmentation method and segmentation device thereof

Abstract

The invention discloses a dermatoscope image segmentation method and a segmentation device thereof, and relates to the technical field of digital image processing, and the segmentation method comprises the steps: adding gray level image information on the basis of an RGB color image of a dermatoscope image, and obtaining a preprocessed dermatoscope image; a lightweight full convolutional neural network is used as a basic framework of an image segmentation model, ResNet50 is used as an encoding stage of the lightweight full convolutional neural network, and in the process, a network structure of ResNet50 and an encoder structure of the lightweight full convolutional neural network are subjected to matching adjustment; aggregating the global information by adopting a dual attention module ina lightweight full convolutional neural network of the image segmentation model to generate a channel attention vector; constructing an adaptive weight loss module; etc. According to the method, thelesion area in the dermatoscope image can be accurately segmented.

Description

technical field [0001] The invention relates to the technical field of digital image processing, in particular to a method and a device for segmenting dermoscopic images. Background technique [0002] Automatic segmentation of lesion regions in dermoscopic images is of great significance for the clinical diagnosis and evaluation of melanoma. Traditional segmentation methods are cumbersome, have poor generalization, and have poor segmentation results for highly similar images between normal and lesion regions. [0003] Moreover, due to the changeable shape of the skin lesion area, the size of the segmentation target and the background is extremely unbalanced, the skin lesion area often presents complex shapes and multi-scale changes, making the segmentation target and the background extremely unbalanced, and it is difficult to segment the precise lesion area . Finally, hair and artificial markers will affect the image quality. Therefore, as far as the current technology is ...

AI Technical Summary

Problems solved by technology

The traditional segmentation method is cumbersome, the generalization is not strong, and the segmentation results of highly similar images between normal and lesion regions are poor

[0003] Moreover, due to the changeable shape of the skin lesion area, the size of the segmentation target and the background is extremely unbalanced, the skin lesion area often presents complex shapes and multi-scale changes, making the segmentation target and the background extremely unbalanced, and it is difficult to segment the precise lesion area

Finally, hair and artificial markers will affect the image quality. Therefore, as far as the current technology is concerned, the accurate segmentation of lesion areas in dermoscopic images is still a big challenge, which needs to be solved as soon as possible to meet the needs of clinical diagnosis and evaluation.

Images