.png){kind=icon}
.png){kind=icon}
Return
Wavelet-Based Adaptive Vision State Space Network for Image Harmonization
.png){kind=icon}
PRE
AI
DOI：10.1007/s11263-026-02981-2.png)
.png){kind=icon}
Original request for help.png){kind=icon}
Share.png){kind=icon}
Save
Abstract
En 中文
Modeling long-range background dependencies while preserving local positional cues of the foreground is crucial for image harmonization. In this paper, we propose a wavelet-based adaptive vision state space network to harmonize local regions by comprehensively exploiting long-range background context. Specifically, we introduce a low-frequency harmonization block to aggregate global context and capture foreground–background interactions. To align the positional information of the foreground object with the local regions of the background, we design an adaptive visual state space module to adaptively learn long-distance dependencies by fusing local spatial relations into the state computation process. In addition, we design a high-frequency modulation block to recover the texture and details of the foreground objects. Experimental results show that the proposed method performs favorably against the state-of-the-art approaches.
Keywords:
Image harmonization
Long-range dependencies
Wavelet
Vision state space network
Journal
IF:
9.3
Papers:
3.9K
Citations:
2.8W