An official website of the United States government
Recent frameworks for image denoising have demonstrated that it can be more productive to recover an image from a smoothed version of some geometric feature of the image rather than denoise the image directly. Improvements can be found both with respect to image quality metrics as well as the preservation of fine details. The challenge in working with this data is that mathematically sound mechanisms developed for handling natural image data do not necessarily apply, and this data itself can be quite ill behaved. In this work we learn both ‘geometric’ or nonlinear higher order features and corresponding regularizers. These approaches show improvement over recent modelbased deep learning (DL) image denoising methods both with respect to image quality metrics as well as the preservation of fine features. Furthermore, the proposed approach for enhancing DL architectures by incorporating geometrically-inspired features is motivated by and has the potential to feed back into mathematically sound models for solving a variety of problems in image processing.
more »
« less
ICON: Learning Regular Maps Through Inverse Consistency
Learning maps between data samples is fundamental. Applications range from representation learning, image translation and generative modeling, to the estimation of spatial deformations. Such maps relate feature vectors, or map between feature spaces. Well-behaved maps should be regular, which can be imposed explicitly or may emanate from the data itself. We explore what induces regularity for spatial transformations, e.g., when computing image registrations. Classical optimization-based models compute maps between pairs of samples and rely on an appropriate regularizer for well-posedness. Recent deep learning approaches have attempted to avoid using such regularizers altogether by relying on the sample population instead. We explore if it is possible to obtain spatial regularity using an inverse consistency loss only and elucidate what explains map regularity in such a context. We find that deep networks combined with an inverse consistency loss and randomized off-grid interpolation yield well behaved, approximately diffeomorphic, spatial transformations. Despite the simplicity of this approach, our experiments present compelling evidence, on both synthetic and real data, that regular maps can be obtained without carefully tuned explicit regularizers, while achieving competitive registration performance.
more »
« less
- Award ID(s):
- 1711776
- PAR ID:
- 10376119
- Date Published:
- Journal Name:
- International Conference on Computer Vision
- Page Range / eLocation ID:
- 3376 to 3385
- Format(s):
- Medium: X
- Sponsoring Org:
- National Science Foundation
More Like this
-
-
We introduce the isoperimetric loss as a regularization criterion for learning the map from a visual representation to a semantic embedding, to be used to transfer knowledge to unknown classes in a zero-shot learning setting. We use a pretrained deep neural network model as a visual representation of image data, a Word2Vec embedding of class labels, and linear maps between the visual and semantic embedding spaces. However, the spaces themselves are not linear, and we postulate the sample embedding to be populated by noisy samples near otherwise smooth manifolds. We exploit the graph structure defined by the sample points to regularize the estimates of the manifolds by inferring the graph connectivity using a generalization of the isoperimetric inequalities from Riemannian geometry to graphs. Surprisingly, this regularization alone, paired with the simplest baseline model, outperforms the state-of-the-art among fully automated methods in zeroshot learning benchmarks such as AwA and CUB. This improvement is achieved solely by learning the structure of the underlying spaces by imposing regularitymore » « less
-
The hippocampus is a crucial brain structure involved in memory formation, spatial navigation, emotional regulation, and learning. An accurate MRI image segmentation of the human hippocampus plays an important role in multiple neuro-imaging research and clinical practice, such as diagnosing neurological diseases and guiding surgical interventions. While most hippocampus segmentation studies focus on using T1-weighted or T2-weighted MRI scans, we explore the use of diffusion-weighted MRI (dMRI), which offers unique insights into the microstructural properties of the hippocampus. Particularly, we utilize various anisotropy measures derived from diffusion MRI (dMRI), including fractional anisotropy, mean diffusivity, axial diffusivity, and radial diffusivity, for a multi-contrast deep learning approach to hippocampus segmentation. To exploit the unique benefits offered by various contrasts in dMRI images for accurate hippocampus segmentation, we introduce an innovative multimodal deep learning architecture integrating cross-attention mechanisms. Our proposed framework comprises a multi-head encoder designed to transform each contrast of dMRI images into distinct latent spaces, generating separate image feature maps. Subsequently, we employ a gated cross-attention unit following the encoder, which facilitates the creation of attention maps between every pair of image contrasts. These attention maps serve to enrich the feature maps, thereby enhancing their effectiveness for the segmentation task. In the final stage, a decoder is employed to produce segmentation predictions utilizing the attention-enhanced feature maps. The experimental outcomes demonstrate the efficacy of our framework in hippocampus segmentation and highlight the benefits of using multi-contrast images over single-contrast images in diffusion MRI image segmentation.more » « less
-
There has been significant recent interest in the use of deep learning for regularizing imaging inverse problems. Most work in the area has focused on regularization imposed implicitly by convolutional neural networks (CNNs) pre-trained for image reconstruction. In this work, we follow an alternative line of work based on learning explicit regularization functionals that promote preferred solutions. We develop the Explicit Learned Deep Equilibrium Regularizer (ELDER) method for learning explicit regularization functionals that minimize a mean-squared error (MSE) metric. ELDER is based on a regularization functional parameterized by a CNN and a deep equilibrium learning (DEQ) method for training the functional to be MSE-optimal at the fixed points of the reconstruction algorithm. The explicit regularizer enables ELDER to directly inherit fundamental convergence results from optimization theory. On the other hand, DEQ training enables ELDER to improve over existing explicit regularizers without prohibitive memory complexity during training. We use ELDER to train several approaches to parameterizing explicit regularizers and test their performance on three distinct imaging inverse problems. Our results show that ELDER can greatly improve the quality of explicit regularizers compared to existing methods, and show that learning explicit regularizers does not compromise performance relative to methods based on implicit regularization.more » « less
-
Aliasing refers to the phenomenon that high frequency signals degenerate into completely different ones after sampling. It arises as a problem in the context of deep learning as downsampling layers are widely adopted in deep architectures to reduce parameters and computation. The standard solution is to apply a lowpass filter (e.g., Gaussian blur) before downsampling. However, it can be suboptimal to apply the same filter across the entire content, as the frequency of feature maps can vary across both spatial locations and feature channels. To tackle this, we propose an adaptive content-aware low-pass filtering layer, which predicts separate filter weights for each spatial location and channel group of the input feature maps. We investigate the effectiveness and generalization of the proposed method across multiple tasks, including image classification, semantic segmentation, instance segmentation, video instance segmentation, and image-to-image translation. Both qualitative and quantitative results demonstrate that our approach effectively adapts to the different feature frequencies to avoid aliasing while preserving useful information for recognition. Code is available at https://maureenzou.github.io/ddac/more » « less
- Free Publicly Accessible Full Text
-
Accepted Manuscript1.0
- Conference Paper:
- https://doi.org/10.1109/ICCV48922.2021.00338
