# SUPER Module for Detail-Sensitive and Cost-Efficient U-Net Variant Decoders

arXiv: https://arxiv.org/abs/2511.11015 · HTML: https://mlsanity.com/paper/2511.11015/

## Key facts

- No citing papers in the map yet (data through 14 Oct 2026).
- Builds on 26 of its 48 references that are in the map.
- First posted on arXiv 14 Nov 2025.

## Builds on (most cited first, 26 of 26)

- [U-Net: Convolutional Networks for Biomedical Image Segmentation](https://mlsanity.com/paper/1505.04597/) (2015)
- [Restormer: Efficient Transformer for High-Resolution Image Restoration](https://mlsanity.com/paper/2111.09881/) (2021)
- [On the Spectral Bias of Neural Networks](https://mlsanity.com/paper/1806.08734/) (2018)
- [Simple Baselines for Image Restoration](https://mlsanity.com/paper/2204.04676/) (2022)
- [AdaBins: Depth Estimation using Adaptive Bins](https://mlsanity.com/paper/2011.14141/) (2020)
- [3D Packing for Self-Supervised Monocular Depth Estimation](https://mlsanity.com/paper/1905.02693/) (2019)
- [clDice -- A Novel Topology-Preserving Loss Function for Tubular Structure Segmentation](https://mlsanity.com/paper/2003.07311/) (2020)
- [Focal Frequency Loss for Image Reconstruction and Synthesis](https://mlsanity.com/paper/2012.12821/) (2020)
- [Multi-level Wavelet-CNN for Image Restoration](https://mlsanity.com/paper/1805.07071/) (2018)
- [Towards Understanding the Spectral Bias of Deep Learning](https://mlsanity.com/paper/1912.01198/) (2019)
- [CARAFE: Content-Aware ReAssembly of FEatures](https://mlsanity.com/paper/1905.02188/) (2019)
- [Boosting Monocular Depth Estimation Models to High-Resolution via Content-Adaptive Multi-Resolution Merging](https://mlsanity.com/paper/2105.14021/) (2021)
- [Attention Attention Everywhere: Monocular Depth Prediction with Skip Attention](https://mlsanity.com/paper/2210.09071/) (2022)
- [iDisc: Internal Discretization for Monocular Depth Estimation](https://mlsanity.com/paper/2304.06334/) (2023)
- [Wavelet Domain Style Transfer for an Effective Perception-distortion Tradeoff in Single Image Super-Resolution](https://mlsanity.com/paper/1910.04074/) (2019)
- [FADE: Fusing the Assets of Decoder and Encoder for Task-Agnostic Upsampling](https://mlsanity.com/paper/2207.10392/) (2022)
- [Catch Missing Details: Image Reconstruction with Frequency Augmented Variational Autoencoder](https://mlsanity.com/paper/2305.02541/) (2023)
- [CascadedGaze: Efficiency in Global Context Extraction for Image Restoration](https://mlsanity.com/paper/2401.15235/) (2024)
- [Addressing Spectral Bias of Deep Neural Networks by Multi-Grade Deep Learning](https://mlsanity.com/paper/2410.16105/) (2024)
- [Towards Building More Robust Models with Frequency Bias](https://mlsanity.com/paper/2307.09763/) (2023)
- [How Convolutional Neural Networks Deal with Aliasing](https://mlsanity.com/paper/2102.07757/) (2021)
- [A Comprehensive Review of U-Net and Its Variants: Advances and Applications in Medical Image Segmentation](https://mlsanity.com/paper/2502.06895/) (2025)
- [Multi-resolution Monocular Depth Map Fusion by Self-supervised Gradient-based Composition](https://mlsanity.com/paper/2212.01538/) (2022)
- [Hybrid-Segmentor: A Hybrid Approach to Automated Fine-Grained Crack Segmentation in Civil Infrastructure](https://mlsanity.com/paper/2409.02866/) (2024)
- [Augmenting Perceptual Super-Resolution via Image Quality Predictors](https://mlsanity.com/paper/2504.18524/) (2025)
- [A Survey of Deep Learning Video Super-Resolution](https://mlsanity.com/paper/2506.03216/) (2025)

## Abstract (arXiv)

Skip-connected U-Net variants are widely used for dense inverse problems, yet their decoders commonly recover resolution through spatial upscaling, which can blur or distort fine structures. Wavelet transforms provide an explicit perfect-reconstruction path, but prior wavelet networks often couple this property to wavelet-specific architectures or fixed output representations. We propose Selectively Suppressed Perfect Reconstruction (SUPER), a wavelet-domain decoder formulation that replaces unstructured spatial refinement with bounded frequency suppression. In its idealized equal-channel form, SUPER reduces to exact reconstruction when the suppression term is zero; in practical U-Net decoders, it provides a projected-subspace reconstruction fallback with learnable task-specific suppression. The resulting SUPER Module is a structurally plug-in decoder block: it replaces the upsampling/refinement stage of an existing U-Net-like decoder, while leaving the encoder and task head unchanged. We evaluate SUPER on monocular depth estimation, thin-crack segmentation, and smartphone image denoising. On iDisc depth estimation, SUPER improves the host model's edge AUC from 26.59% to 34.45% while reducing decoder MACs from 12.69G to 4.36G. On FACS-Net crack segmentation, SUPER improves average IoU and the extreme 0-2px crack regime. On SIDD denoising, where high-frequency enhancement is less directly rewarded, SUPER preserves PSNR/SSIM while reducing decoder MACs by 66.6%. These results support SUPER as a practical wavelet-domain suppression module for improving the observed detail-cost trade-off of U-Net-like decoders.