There are many super-resolution techniques which show their potential in achieving super-resolution images which attempt to reach the quality of HR images. This report provides the detailed overview of most important research articles starting from such traditional methods as interpolation to deep neural networks which have achieved great success in single image super-resolution (SISR). As the idea of skip connections promotes better performance, UNet was implemented and trained with pixel-wise losses l1 loss, L2 loss, and Smooth L1-loss on SISR dataset of ultrasound medical images collected for this research as a baseline solution to SISR. In order to overcome some limitations as simple, one-step fusion in UNet, as well as to improve visual SISR results, we propose some modifications to this baseline as a further direction of our research - SISR with the Generative Adversarial Network, and Deep Layer Aggregation.
-
Notifications
You must be signed in to change notification settings - Fork 0
High resolution (HR) is desirable in many areas, in particular, in medical imaging. Since such images are an important method to find certain diseases, the high resolution should improve the success rate of correct diagnosis. There are many super-resolution techniques which show their potential in achieving super-resolution images which attempt …
MeazaEyakem1/Application-of-AI-in-Medical-Imaging-
This commit does not belong to any branch on this repository, and may belong to a fork outside of the repository.
Folders and files
Name | Name | Last commit message | Last commit date | |
---|---|---|---|---|
Repository files navigation
About
High resolution (HR) is desirable in many areas, in particular, in medical imaging. Since such images are an important method to find certain diseases, the high resolution should improve the success rate of correct diagnosis. There are many super-resolution techniques which show their potential in achieving super-resolution images which attempt …
Resources
Stars
Watchers
Forks
Releases
No releases published
Packages 0
No packages published