OracleZoom:基于参考约束的递归图像超分辨率,突破真实标签极限
OracleZoom: Reference-Constrained Recursive Image Super-Resolution Beyond Ground Truth
AI 导读
arXiv 论文提出 OracleZoom 递归图像超分辨率方法,将最后一次有真实标签(如 $4\times$)的图像区域作为训练信号,通过跨尺度损失把 $16\times$ 预测投影回真实标签,并用无参考质量目标与 latent prior 补充细节、抑制幻觉。
正文 · 原文
Abstract:Recursive image Super-Resolution (SR) reaches extreme magnification by feeding each prediction back into the same model, like zooming into an image again and again. Real ground truth for deep zooms is impractical to capture (a full $256\times$ target would be a 52 GB image), so predictions at deeper zooms have no supervision. Yet the last ground-truth image still contains the region that every deeper zoom magnifies. OracleZoom turns this region into a training signal. Inspired by on-policy self-distillation, the model trains on its own prediction for one recursive step ($4\times \rightarrow 16\times$), and a cross-scale loss projects each $16\times$ prediction back to the last ground truth at $4\times$. This projection can verify only coarse content, so a no-reference quality objective guides the remaining fine detail, and a latent prior stops this objective from adding hallucinated artifacts. On seven datasets, OracleZoom achieves the highest CLIPIQA at every zoom depth (mean 0.713 vs. 0.621 for Chain-of-Zoom), with the largest gains at the deepest zooms, even though training stops at $16\times$ and optimizes a different quality metric. Pooled over two ground-truth-anchored judges, this added detail comes at the same hallucination rate as Chain-of-Zoom at $16\times$ and a lower rate at $4\times$, and removing the cross-scale loss raises hallucination at $16\times$.
| Comments: | Project page: this https URL |
| Subjects: | Computer Vision and Pattern Recognition (cs.CV); Artificial Intelligence (cs.AI); Computation and Language (cs.CL); Machine Learning (cs.LG) |
| Cite as: | arXiv:2609.06490 [cs.CV] |
| (or arXiv:2609.06490v2 [cs.CV] for this version) | |
| https://doi.org/10.48550/arXiv.2609.06490 arXiv-issued DOI via DataCite |
Submission history
From: Shubhashis Roy Dipta [view email]
[v1]
Sun, 6 Sep 2026 09:06:27 UTC (42,313 KB)
[v2]
Wed, 18 Nov 2026 05:32:39 UTC (32,084 KB)
来源:arXiv cs.AI · arxiv.org