Investigation of Denoising Techniques for Removal of Hair and Noise from Dermoscopic Images

Skin cancer detection is a complicated process where skin images are processed to detect and classify skin diseases. Dermoscopy image analysis of skin lesions are automatically provided by computer-aided systems. The issue and challenge in conventional research are that dermoscopy images captured by dermoscopic devices contain noise, which reduces the accuracy of automated computer-aided-system. But it has been observed that skin cancer detection could be improved by eliminating noise and hair artifacts. However, a noise reduction technique might be implemented to guarantee the best possible picture quality by measurements of ISNR, SSIM, and MS E. The present research study is focused on preprocessing of skin images that are image scaling, hair removal, and noise removal to resolve the issues related to noise and accuracy that have been found in conventional research. The outcomes have been examined numerically and graphically to compare the abilities of the systems. The objective of the work is to make the skin images more understandable to deep learning mechanisms during classification operation. Preprocessing of the image is considering image resizing, hair removal, and noise removal. Thus, the proposed wo rk is supposed to provide a better contribution during skin cancer image detection.