基于统计特征库的高分卫星影像匀色方法

打开文本图片集

DOI：10.19981/j.CN23-1581/G3.2026.11.008

中图分类号：P237

文献标志码：A

文章编号：2095-2945（2026）11-0036-06

Abstract： High-resolution remote sensing satellite imagery is prone to uneven color distribution in mosaicking due to imaging factors， where color balancing becomes a critical step. To address the bottleneck of massive reference image storage and management， this study proposes an efficient color balancing method based on a statistical feature reference database. By leveraging Gaussian pyramids or wavelet transforms to separate high- and low-frequency image components， the method focuses on low-frequency component balancing. A lightweight reference database storing only statistical features （e.g.， mean， variance） of reference images replaces traditional full-image storage. Experimental results demonstrate： The reference database significantly reduces storage requirements while enhancing portability and management efficiency; for cloud-free images， the proposed CIM （Cloud-Insensitive Method） achieves superior visual and quantitative performance compared to histogram matching， Wallis filtering， and IR-MAD， effectively preserving details; and for cloudy images， integrating SVM-based cloud detection with CIM ensures robust results. This approach provides a practical solution for high-quality， large-scale color balancing of high-resolution satellite imagery in operational workflows.

Keywords： color balancing method; reference image library; CIM; cloud detection; high-resolution remote sensing satellite

影像匀色处理的实质是实现不同影像的辐射归一化，是高分辨率遥感卫星影像处理的关键环节。（剩余8174字）