地球信息科学学报

地球信息科学学报 ›› 2011, Vol. 13 ›› Issue (4): 556-561.doi: 10.3724/SP.J.1047.2011.00556

• 遥感技术与应用 • 上一篇    下一篇

不同景观特征遥感图像融合的最佳分解层数选取分析

郭冠华1, 陈颖彪1, 吴志峰1,2, 魏建兵2   

  1. 1. 广州大学 地理科学学院,广州 510006;
    2. 广东省生态环境与土壤研究所,广州 510650
  • 收稿日期:2011-03-23 修回日期:2011-06-07 出版日期:2011-08-25 发布日期:2011-08-23
  • 基金资助:

    中国科学院资源与环境信息系统国家重点实验室开放研究基金(A0710,2010KF0006SA);国家自然科学基金项目(40871229);广东省自然科学基金项目(9151065003000000)。

Optimal Number of Decomposition Levels for Fusion of Different Landscape Images

GUO Guanhua1, CHEN Yingbiao1, WU Zhifeng1,2, WEI Jianbing2   

  1. 1. School of Geographical Sciences, Guangzhou University, Guangzhou 510006, China;
    2. Guangdong Institute of Eco-environment and Soil Sciences, Guangzhou 510650, China
  • Received:2011-03-23 Revised:2011-06-07 Online:2011-08-25 Published:2011-08-23

PDF (PC)

赞

被引次数

13

摘要： IHS和小波变换相结合的融合方法是一种高效的融合算法。影响该算法性能的因素有很多,其中分解层数的选取对融合图像质量有重要影响,故针对不同景观特征影像选取最佳分解层数的问题有待深入探讨。本文以SPOT全色影像和TM多光谱影像,选取信息熵、平均梯度和相关系数3个质量指标,就不同景观特征影像对小波分解层数的响应问题开展了研究。结果表明:融合图像质量与原影像地物景观特征有密切关系;不同景观融合图像信息熵在分解层数下表现出较强的变异,相关系数变异较弱,就单一景观影像而言,林地景观对分解层数更为敏感;各质量指标的变异规律差异明显,不同景观影像临界层数也各异,可根据其变异曲线特征针对不同景观特征影像确定最佳分解层数。

关键词: 影像融合, 景观特征, 小波变换, 最佳层数

Abstract: The joint use of IHS and wavelet transforms is a popular fusion method to incorporate multi-spectral remote data and high-resolution panchromatic data. However, some important factors have been directly neglected when this method is adopted. In these factors, number of decomposition levels is the key to influence the effect of the fusion images, and the problem about how to choose the optimal number of decomposition level according to remote images with different landscape characteristics should be focused on. In this paper, SPOP panchromatic data with 2.5×2.5m and TM mutli-spectral data with 30×30 m were used, entropy, average gradient and correlation coefficients were calculated as the evaluation indices for fusion images. The object of this paper was to explore the response of fusion images performances from different landscape characteristics to the choice of number of decomposition levels, and according to those results from the fusion treatments we found the optimal number of the levels to the given remote images. The results showed that the performances of fusion images had a close relationship with land-cover information and landscape types included in the remote images obviously. Entropy of different landscape images displayed very strong variance with changing decomposition levels, but those with correlation coefficients were slight. Comparing with other types of remote images, forest landscape images showed the most sensitive relationship with number of decomposition levels. Different indices exhibited different characteristics to decomposition levels, and images with different landscape information had their specific critical decomposition levels. According to the analysis given above, the optimal number of the levels to a given remote image could be confirmed. For example, five was the optimal number of the levels of urban landscape image. With this choice, fusion image exhibited the best performance in three evaluation indices. This paper tries to provide helpful information when we use the integration of wavelet transforms and IHS as the method in multiple sources remotely sensed data fusion.

Key words: wavelet transform, images fusion, landscape characteristics, optimal number of levels