AlignVAR: Towards Globally Consistent Visual Autoregression for Image Super-Resolution

Fuente: arXiv
Salvato in:
Dettagli Bibliografici
Autori principali: Liu, Cencen, Zhang, Dongyang, Yin, Wen, Wang, Jielei, Li, Tianyu, Guo, Ji, Jiang, Wenbo, Wang, Guoqing, Lu, Guoming
Natura: Preprint
Pubblicazione: 2026
Soggetti:
Accesso online:
Tags: Aggiungi Tag
Nessun Tag, puoi essere il primo ad aggiungerne!!
_version_ 1866911485708468224
author Liu, Cencen
Zhang, Dongyang
Yin, Wen
Wang, Jielei
Li, Tianyu
Guo, Ji
Jiang, Wenbo
Wang, Guoqing
Lu, Guoming
author_facet Liu, Cencen
Zhang, Dongyang
Yin, Wen
Wang, Jielei
Li, Tianyu
Guo, Ji
Jiang, Wenbo
Wang, Guoqing
Lu, Guoming
contents Visual autoregressive (VAR) models have recently emerged as a promising alternative for image generation, offering stable training, non-iterative inference, and high-fidelity synthesis through next-scale prediction. This encourages the exploration of VAR for image super-resolution (ISR), yet its application remains underexplored and faces two critical challenges: locality-biased attention, which fragments spatial structures, and residual-only supervision, which accumulates errors across scales, severely compromises global consistency of reconstructed images. To address these issues, we propose AlignVAR, a globally consistent visual autoregressive framework tailored for ISR, featuring two key components: (1) Spatial Consistency Autoregression (SCA), which applies an adaptive mask to reweight attention toward structurally correlated regions, thereby mitigating excessive locality and enhancing long-range dependencies; and (2) Hierarchical Consistency Constraint (HCC), which augments residual learning with full reconstruction supervision at each scale, exposing accumulated deviations early and stabilizing the coarse-to-fine refinement process. Extensive experiments demonstrate that AlignVAR consistently enhances structural coherence and perceptual fidelity over existing generative methods, while delivering over 10x faster inference with nearly 50% fewer parameters than leading diffusion-based approaches, establishing a new paradigm for efficient ISR.
format Preprint
id arxiv_https___arxiv_org_abs_2603_00589
institution arXiv
publishDate 2026
record_format arxiv
spellingShingle AlignVAR: Towards Globally Consistent Visual Autoregression for Image Super-Resolution
Liu, Cencen
Zhang, Dongyang
Yin, Wen
Wang, Jielei
Li, Tianyu
Guo, Ji
Jiang, Wenbo
Wang, Guoqing
Lu, Guoming
Computer Vision and Pattern Recognition
Artificial Intelligence
Visual autoregressive (VAR) models have recently emerged as a promising alternative for image generation, offering stable training, non-iterative inference, and high-fidelity synthesis through next-scale prediction. This encourages the exploration of VAR for image super-resolution (ISR), yet its application remains underexplored and faces two critical challenges: locality-biased attention, which fragments spatial structures, and residual-only supervision, which accumulates errors across scales, severely compromises global consistency of reconstructed images. To address these issues, we propose AlignVAR, a globally consistent visual autoregressive framework tailored for ISR, featuring two key components: (1) Spatial Consistency Autoregression (SCA), which applies an adaptive mask to reweight attention toward structurally correlated regions, thereby mitigating excessive locality and enhancing long-range dependencies; and (2) Hierarchical Consistency Constraint (HCC), which augments residual learning with full reconstruction supervision at each scale, exposing accumulated deviations early and stabilizing the coarse-to-fine refinement process. Extensive experiments demonstrate that AlignVAR consistently enhances structural coherence and perceptual fidelity over existing generative methods, while delivering over 10x faster inference with nearly 50% fewer parameters than leading diffusion-based approaches, establishing a new paradigm for efficient ISR.
title AlignVAR: Towards Globally Consistent Visual Autoregression for Image Super-Resolution
topic Computer Vision and Pattern Recognition
Artificial Intelligence
url https://arxiv.org/abs/2603.00589