Image Enhancment Hair Removal in Dermoscopy Images Using Dual Swin Autoencoder

Skin cancer requires effective treatment for better survival rates especially in the case of melanoma. To reduce the image noise and removing hair the conventional method takes high computational complexity and decayed accuracy. In order to tackle this issue, an innovative dual Swin autoencoder has been proposed in this paper. This autoencoder utilizes Swin Transformer blocks to effectively eliminate noise from images and get rid of unwanted hair, all while maintaining important characteristics of the skin and lesions. Experimental results demonstrate that our model outperforms conventional techniques, with PSNR values up to 81.71 dB and SSIM up to 0.99, effectively eliminating hair and retaining important skin features. This approach enhances the accuracy of melanoma diagnosis in dermoscopy images.