eduzhai > Applied Sciences > Engineering >

TextureWGAN Texture Preserving WGAN with MLE Regularizer for Inverse Problems

  • king
  • (0) Download
  • 20210506
  • Save

... pages left unread,continue reading

Document pages: 4 pages

Abstract: Many algorithms and methods have been proposed for inverse problemsparticularly with the recent surge of interest in machine learning and deeplearning methods. Among all proposed methods, the most popular and effectivemethod is the convolutional neural network (CNN) with mean square error (MSE).This method has been proven effective in super-resolution, image de-noising,and image reconstruction. However, this method is known to over-smooth imagesdue to the nature of MSE. MSE based methods minimize Euclidean distance for allpixels between a baseline image and a generated image by CNN and ignore thespatial information of the pixels such as image texture. In this paper, weproposed a new method based on Wasserstein GAN (WGAN) for inverse problems. Weshowed that the WGAN-based method was effective to preserve image texture. Italso used a maximum likelihood estimation (MLE) regularizer to preserve pixelfidelity. Maintaining image texture and pixel fidelity is the most importantrequirement for medical imaging. We used Peak Signal to Noise Ratio (PSNR) andStructure Similarity (SSIM) to evaluate the proposed method quantitatively. Wealso conducted first-order and second-order statistical image texture analysisto assess image texture.

Please select stars to rate!

         

0 comments Sign in to leave a comment.

    Data loading, please wait...
×