高光谱影像具有“谱像合一”的特征, 因而可以从谱和像两个角度对高光谱影像进行分形分析。本文把计 算表面分形维的双层地毯法用于计算Hyperion 高光谱曲线的分形维, 根据每一像元的高光谱曲线计算出每一尺度 的上分形信号和下分形信号, 并对分形信号值及分形特征尺度进行分析。结果表明对于不同地物其分形信号在某些 特定尺度(分形特征尺度)有较大的差异, 而且对于Hyperion 高光谱影像, 下分形信号可更有效地突出地物特征。文 章最后简要讨论了上分形和下分形、起始计算尺度以及原始采样点数目对分形特征尺度的影响。

A fractal feature may be analyzed spectrally and spatially due to the “combination of spectrum and image” character inhered in hyperspectral images. The blanket method used for computing fractal dimension of surface is transferred to calculating fractal dimension of curves of Hyperion image. Up and down fractal signature for each scale is calculated based on the hyperspectral curve of each pixel. The fractal signature and their corresponding scale are analyzed. The result shows that fractal signature changes obviously at some special scales for different targets. As for the Hyperion data, the down fractal signature is more salient for effectively discriminating targets. A brief discussion about the relation of fractal feature scale with up and down fractal signature, initial calculating scale and number of samples is presented at the end of this paper.