Abstract:
Accurate diagnostics require high-resolution MR images. Hardware and processing limits image resolution. Deep learning has produced impressive image enhancement/super-resolution results recently.
We propose a new regularized network that uses image priors—a low-rank structure and a sharpness prior—to improve deep MR image super-resolution (SR). We then incorporate these priors in an analytically tractable manner and create a novel prior-guided network architecture that performs super-resolution.
Differentiable rank approximations help the low rank prior since the rank is not a differentiable function of the image matrix (and thus the network parameters). We demonstrate that a fixed feedback layer at the network’s output can implement the Laplacian’s variance to emphasize sharpness.
We add sharpness-optimized training data-driven filters to the fixed feedback (Laplacian) layer. The proposed prior guided network improves SNR/image quality in publicly available MR brain image databases. Our priors are on output images, so the proposed method can be used with many network architectures to improve performance.
Note: Please discuss with our team before submitting this abstract to the college. This Abstract or Synopsis varies based on student project requirements.
Did you like this final year project?
To download this project Code with thesis report and project training... Click Here
