# Untrained Perceptual Loss for image denoising of line-like structures in MR images

arXiv: https://arxiv.org/abs/2411.05884 · HTML: https://mlsanity.com/paper/2411.05884/

## Key facts

- No citing papers in the map yet (data through 14 Oct 2026).
- Builds on 23 of its 55 references that are in the map.
- First posted on arXiv 8 Nov 2024.

## Builds on (most cited first, 23 of 23)

- [Delving Deep into Rectifiers: Surpassing Human-Level Performance on ImageNet Classification](https://mlsanity.com/paper/1502.01852/) (2015)
- [Perceptual Losses for Real-Time Style Transfer and Super-Resolution](https://mlsanity.com/paper/1603.08155/) (2016)
- [Photo-Realistic Single Image Super-Resolution Using a Generative Adversarial Network](https://mlsanity.com/paper/1609.04802/) (2016)
- [Restormer: Efficient Transformer for High-Resolution Image Restoration](https://mlsanity.com/paper/2111.09881/) (2021)
- [Beyond a Gaussian Denoiser: Residual Learning of Deep CNN for Image Denoising](https://mlsanity.com/paper/1608.03981/) (2016)
- [Making Convolutional Networks Shift-Invariant Again](https://mlsanity.com/paper/1904.11486/) (2019)
- [Residual Non-local Attention Networks for Image Restoration](https://mlsanity.com/paper/1903.10082/) (2019)
- [Photographic Image Synthesis with Cascaded Refinement Networks](https://mlsanity.com/paper/1707.09405/) (2017)
- [Generating Images with Perceptual Similarity Metrics based on Deep Networks](https://mlsanity.com/paper/1602.02644/) (2016)
- [Synthesizing the preferred inputs for neurons in neural networks via deep generator networks](https://mlsanity.com/paper/1605.09304/) (2016)
- [MR Image Denoising and Super-Resolution Using Regularized Reverse Diffusion](https://mlsanity.com/paper/2203.12621/) (2022)
- [SUNet: Swin Transformer UNet for Image Denoising](https://mlsanity.com/paper/2202.14009/) (2022)
- [Multi-Modal Transformer for Accelerated MR Imaging](https://mlsanity.com/paper/2106.14248/) (2021)
- [Super-Resolution with Deep Convolutional Sufficient Statistics](https://mlsanity.com/paper/1511.05666/) (2015)
- [SDnDTI: Self-supervised deep learning-based denoising for diffusion tensor MRI](https://mlsanity.com/paper/2111.07220/) (2021)
- [A Powerful Generative Model Using Random Weights for the Deep Image Representation](https://mlsanity.com/paper/1606.04801/) (2016)
- [Generic Perceptual Loss for Modeling Structured Output Dependencies](https://mlsanity.com/paper/2103.10571/) (2021)
- [Denoising of 3-D Magnetic Resonance Images Using a Residual Encoder-Decoder Wasserstein Generative Adversarial Network](https://mlsanity.com/paper/1808.03941/) (2018)
- [MRI denoising using Deep Learning and Non-local averaging](https://mlsanity.com/paper/1911.04798/) (2019)
- [Perceptual cGAN for MRI Super-resolution](https://mlsanity.com/paper/2201.09314/) (2022)
- [Deep learning-based reconstruction of highly accelerated 3D MRI](https://mlsanity.com/paper/2203.04674/) (2022)
- [Imaging transformer for MRI denoising with the SNR unit training: enabling generalization across field-strengths, imaging contrasts, and anatomy](https://mlsanity.com/paper/2404.02382/) (2024)
- [DARCS: Memory-Efficient Deep Compressed Sensing Reconstruction for Acceleration of 3D Whole-Heart Coronary MR Angiography](https://mlsanity.com/paper/2402.00320/) (2024)

## Abstract (arXiv)

In the acquisition of Magnetic Resonance (MR) images shorter scan times lead to higher image noise. Therefore, automatic image denoising using deep learning methods is of high interest. MR images containing line-like structures such as roots or vessels yield special characteristics as they display connected structures and yield sparse information. For this kind of data, it is important to consider voxel neighborhoods when training a denoising network. In this paper, we translate the Perceptual Loss to 3D data by comparing feature maps of untrained networks in the loss function as done previously for 2D data. We tested the performance of untrained Perceptual Loss (uPL) on 3D image denoising of MR images displaying brain vessels (MR angiograms - MRA) and images of plant roots in soil. We investigate the impact of various uPL characteristics such as weight initialization, network depth, kernel size, and pooling operations on the results. We tested the performance of the uPL loss on four Rician noise levels using evaluation metrics such as the Structural Similarity Index Metric (SSIM). We observe, that our uPL outperforms conventional loss functions such as the L1 loss or a loss based on the Structural Similarity Index Metric (SSIM). The uPL network's initialization is not important, while network depth and pooling operations impact denoising performance. E.g. for both datasets a network with five convolutional layers led to the best performance while a network with more layers led to a performance drop. We also find that small uPL networks led to better or comparable results than using large networks such as VGG. We observe superior performance of our loss for both datasets, all noise levels, and three network architectures. In conclusion, for images containing line-like structures, uPL is an alternative to other loss functions for 3D image denoising.