RECORD
WAVE: Wavelet-Controlled Coarse-to-Fine Guidance for Depth Super-Resolution
Computer Vision and Pattern Recognition · Submitted 26 Aug 2026
abstract compiled · 2026-09-23 · no language-model review
The authors' main claim Extensive experiments across multiple benchmarks demonstrate that WAVE matches or outperforms existing methods, with the largest gains at high upsampling factors, where low-resolution depth contains the least structure.
The compiled record
The paper's own sentences, arranged. Each label is drawn from a closed vocabulary; the number beside it is the probability the classifier gave that label. Numbers in the tables are the authors', checked by code. No sentence in this record was written by a model.
Open the compiled record
The abstract, sentence by sentence
Guided depth super-resolution (GDSR) typically extracts RGB guidance features through convolutional hierarchies, inheriting their fine-to-coarse bias. Thus, low-level spatial cues surface in early layers, leaving the deeper layers to suppress those that do not correspond to true depth boundaries, which risks artifacts and blurred edges. The same fine-to-coarse bias persists in semantics-based methods that consume low-level tokens early and global tokens late. We present WAVE, which introduces a multi-level discrete wavelet transform (ML-DWT) as an explicit and interpretable feature-control mechanism, enabling a coarse-to-fine reconstruction by consuming sub-bands and semantic tokens in reverse of their generation order. WAVE further exploits these sub-bands to treat high- and low-frequency content separately, filtering at its source the misleading RGB color and texture cues that often lead to blurred boundaries and artifacts, offering an intuitive alternative to the suppression learned implicitly by an opaque network. WAVE separates structure and detail reconstruction into dedicated modules that: i) model interactions within and across wavelet sub-bands, depth features, and semantic priors, ii) apply semantic gating to the high-frequency bands, and iii) fuse modalities through an invertible coupling mechanism that prevents collapse onto a single modality. Extensive experiments across multiple benchmarks demonstrate that WAVE matches or outperforms existing methods, with the largest gains at high upsampling factors, where low-resolution depth contains the least structure.
background contribution method result interpretation main claim
| # | sentence | role | the authors' own work | states the main conclusion | strength |
|---|---|---|---|---|---|
| s0 | Guided depth super-resolution (GDSR) typically extracts RGB guidance features through convolutional hierarchies, inheriting their fine-to-coarse bias. | background 1.00 | 0.05 | 0.03 | asserted 0.99 |
| s1 | Thus, low-level spatial cues surface in early layers, leaving the deeper layers to suppress those that do not correspond to true depth boundaries, which risks artifacts and blurred edges. | background 0.94 | 0.07 | 0.04 | asserted 0.66 |
| s2 | The same fine-to-coarse bias persists in semantics-based methods that consume low-level tokens early and global tokens late. | background 0.97 | 0.05 | 0.04 | asserted 1.00 |
| s3 | We present WAVE, which introduces a multi-level discrete wavelet transform (ML-DWT) as an explicit and interpretable feature-control mechanism, enabling a coarse-to-fine reconstruction by consuming sub-bands and semantic tokens in reverse of their generation order. | contribution 0.77 | 0.94 | 0.29 | asserted 1.00 |
| s4 | WAVE further exploits these sub-bands to treat high- and low-frequency content separately, filtering at its source the misleading RGB color and texture cues that often lead to blurred boundaries and artifacts, offering an intuitive alternative to the suppression learned implicitly by an opaque network. | method 0.81 | 0.84 | 0.14 | asserted 0.99 |
| s5 | WAVE separates structure and detail reconstruction into dedicated modules that: i) model interactions within and across wavelet sub-bands, depth features, and semantic priors, ii) apply semantic gating to the high-frequency bands, and iii) fuse modalities through an invertible coupling mechanism that prevents collapse onto a single modality. | method 0.97 | 0.91 | 0.07 | asserted 0.99 |
| s6 | Extensive experiments across multiple benchmarks demonstrate that WAVE matches or outperforms existing methods, with the largest gains at high upsampling factors, where low-resolution depth contains the least structure. | result 1.00 | 0.96 | 0.69 | asserted 1.00 |
What was examined, and what was not
- arXiv listing abstract v1
- full text
- figures and tables
- references and citing sentences
- relations to other papers
- P abstract
- S sections
- T body sentences
- C citations
- R cross-paper relations
The paper
Original abstract
Guided depth super-resolution (GDSR) typically extracts RGB guidance features through convolutional hierarchies, inheriting their fine-to-coarse bias. Thus, low-level spatial cues surface in early layers, leaving the deeper layers to suppress those that do not correspond to true depth boundaries, which risks artifacts and blurred edges. The same fine-to-coarse bias persists in semantics-based methods that consume low-level tokens early and global tokens late. We present WAVE, which introduces a multi-level discrete wavelet transform (ML-DWT) as an explicit and interpretable feature-control mechanism, enabling a coarse-to-fine reconstruction by consuming sub-bands and semantic tokens in reverse of their generation order. WAVE further exploits these sub-bands to treat high- and low-frequency content separately, isolating the misleading RGB color and texture cues before any feature is learned so that their admission can be controlled explicitly, offering an intuitive alternative to the suppression learned implicitly by an opaque network. WAVE separates structure and detail reconstruction into dedicated modules that: i) model interactions within and across wavelet sub-bands, depth features, and semantic priors, ii) apply semantic gating to the high-frequency bands, and iii) fuse modalities through an invertible coupling mechanism that prevents collapse onto a single modality. Extensive experiments across multiple benchmarks demonstrate that WAVE matches or outperforms existing methods, with the largest gains at high upsampling factors, where low-resolution depth contains the least structure.
Human review
How the literature uses it
What this paper does with the arXiv papers it cites
What it cites · 13 sentences without a confident label
Uses the method or tool of 2 papers
-
SToRA: We use DINOv3 [ref] (ViT-B/16) to extract patch tokens ${τ_l}_l=1^4$ from 4 layers (2, 5, 9, 11), each capturing a different level of semantic abstraction, supplying semantic priors from $I$ as a derived auxiliary guiding modality.
DINOv3 · WAVE - Coarse-to-Fine Reconstruction · 1.00
-
For optimization we use modified $L_1$ loss [ref] .
Implicit Neural Representation-Based Continuous Single Image Super-Resolution: An Empirical Benchmark · Boundary Refinement and Optimization · 0.97
Cites as background 2 papers
-
Vision foundation models such as DINO [ref] and SAM [ref] have become a source of knowledge distillation for improving different vision tasks [ref] .
DINOv3 · Introduction · 0.91
-
Vision foundation models such as DINO [ref] and SAM [ref] have become a source of knowledge distillation for improving different vision tasks [ref] .
Depth Anything 3: Recovering the Visual Space from Any Views · Introduction · 0.91
Discussion
Know something about this paper? Leave a comment, or highlight any passage above to comment on it.