Interactive Attention Network Based on Quaternion and Transformer for Hyperspectral Image Classification
Year: 2026
Type: conference-paper
Abstract: The algebraic structure and operational properties of quaternions can preserve more information and enhance information interaction. Thus, this paper proposes a quaternion-Transformer interactive attention network (QTIAN) for hyperspectral image (HSI) classification. First, data is mapped to the quaternion space through data preprocessing and data transformation, which provides a more compact representation for HSI and restricts the degrees of freedom of the model. Next, a quaternion interactive attention module is designed to realize the interaction and fusion of global and local features. Finally, the quaternion Transformer with rotational positional embedding is utilized to enhance the global modeling capability of the network and the richness of information. Experimental results demonstrate the effectiveness of the proposed algorithm.
Institutions: Wuhan Textile University, Wuhan College
Language: English
Subfield: Computer Vision and Pattern Recognition
Field: Computer Science
Domain: Physical Sciences
Open Access status: hybrid
OpenAlex ID: W7203931372