Search papers, labs, and topics across Lattice.
This paper introduces a novel pan-sharpening method based on the RWKV architecture, addressing its semantic-agnostic scanning and positional bias. They propose a Multigrain-aware Semantic Prototype Scanning strategy using locality-sensitive hashing for context-aware token reordering, complemented by a tri-token prompting mechanism for enhanced semantic priors and noise suppression. The method also incorporates an invertible Q-Shift operation for efficient feature transformation, achieving state-of-the-art results in pan-sharpening.
Ditch raster scanning: this new RWKV-based pan-sharpening method uses semantic prototypes for context-aware token reordering, boosting performance.
In this work, we propose a Multigrain-aware Semantic Prototype Scanning paradigm for pan-sharpening, built upon a high-order RWKV architecture and a tri-token prompting mechanism derived from semantic clustering. Specifically, our method contains three key components: 1) Multigrain-aware Semantic Prototype Scanning. Although RWKV offers a efficient linear-complexity alternative to Transformers, its conventional bidirectional raster scanning is still semantic-agnostic and prone to positional bias. To address this issue, we introduce a semantic-driven scanning strategy that leverages locality-sensitive hashing to group semantically related regions and construct multi-grain semantic prototypes, enabling context-aware token reordering and more coherent global interaction. 2) Tri-token Prompt Learning. We design a tri-token prompting mechanism consisting of a global token, cluster-derived prototype tokens, and a learnable register token. The global and prototype tokens provide complementary semantic priors for RWKV modeling, while the register token helps suppress noisy and artifact-prone intermediate representations. 3) Invertible Q-Shift. To counteract spatial details, we apply center difference convolution on the value pathway to inject high-frequency information, and introduce an invertible multi-scale Q-shift operation for efficient and lossless feature transformation without parameter-heavy receptive field expansion. Experimental results demonstrate the superiority of our method.
