首页 >  2008, Vol. 12, Issue (6) : 1993-2002

摘要

全文摘要次数: 4155 全文下载次数: 3120
引用本文:

DOI:

10.11834/jrs.200806129

收稿日期:

修改日期:

PDF Free   HTML   EndNote   BibTeX
遥感影像空间尺度上推方法的评价
1.武汉大学测绘遥感信息工程国家重点实验室,湖北 武汉 430079;2.南京大学地理信息科学系,江苏 南京 210093
摘要:

首先分析了几种常用的影像质量评价指标在遥感数据空间尺度上推方法评价中的不足,同时提出了已有的关于空间尺度上推方法在评价思路上的不妥之处.引入了空间分辨率和SSIM(structural SIMilarity)2个遥感影像质量评价指标,给出了新的评价思路,并在此基础上对5种遥感数据空间尺度转换方法进行了评价.在本实验中,空间分辨率和SSIM 2个遥感影像质量评价指标一致表明,Bilinear和Bicubic方法的结果影像能够更好地接近目标空间尺度下的影像特征.实验表明,依据新的评价思路,采用空间分辨率和SSIM 2个评价指标进行空间尺度上推方法表现出较强的有效性和优越性.

Comparing the Effect of Aggregation Methods for Remote Sensing Image
Abstract:

Spatial data aggregation iswidely practiced for“scaling-up”environmental analyses andmodeling from local to regional orglobalscales. There is a growing literature reporting on the effectsofdata aggregation under the general topic of the scale effects.Themeasurements, such asmean, median, variance,RMSE,etc.,are used to evaluate the effect of differentaggregationmethods in the studies. However,thesemeasurementsbased on statisticshave not taken into account the spatial pattern of the remote sensing mi age. Despite this, the original mi age is setas the reference data and the aggregation method which can retain statistical characteristics of the originaldata better than the others is considered the desira- blemethod. For scientific inquiry, aggregating data to a coarser resolution is often performed, because certain spatialpat-ternswillnotbe revealed until the data are presented ata coarser scale. It ismore reasonable that the mi age scanned directly from sensor at the same scalewith the result resolution of aggregation operation used as reference data.The objective of this study is to compare and evaluate the effect of five aggregationmethods based on newmeasure-ments and new comparing thought line. Spatial resolution and Structural SI Milarity(SSIM) are introduced as two new measurements to evaluate the effects of aggregationmethods. In the five aggregation methods, except for Arithmetic Average Variability-Weighted (AAVW), Average, Bicubic, Bilinear andNearestare commonly used for aggregating remotely sensed data. The AAVW method is a new aggregationmethod based on spatial variability-weighted principle.Spatial resolution is a basic attribute of mi age data. The technical spatial resolution ofup-scaled mi age can be specified precisely and explicitlyby the resolution oforiginal mi age and the aggregation levels. The aggregation effects are eval-uated by differences between the resolution of aggregated mi ages and the technical spatial resolution. In this study, the spatial resolutionwas retrieved by using linear diffuse function. Themotivation of SSIM is to find amore directway to compare the structures of the reference and the distorted mi ages. Natural mi age signals are highly structured: theirpixels exhibit strong dependencies, especially when they are spatially proxmi ate, and these dependencies carry mi portant information about the structure of the objects in the visual scene. SSIM defines the structural information in an mi age as those attributes that represent the structure ofobjects in the scene, independent of the average luminance and contrast.Two study sites from the Xinjiang Province, Chinawere used: TheUrumchi international airport and Tianshan Moun- tain. Foreach study site, the TM and ETM+ mi ageswere acquired. Themulti-spectral mi ages of ETM+weremerged with its panchromatic mi age. Thus, the resolution ofETM+multi-spectral mi ageswas enhanced to 15 m. Themerged ETM+multi-spectral miageswere then aggregated by the five aggregation method smentioned above using 2-by-2-pixelwindow size. The up-scaled mi ages in Urumchi international airportwere selected for the expermi ent based on spatial resolution measurement. SSIM was used in comparison of effects of the five aggregation methods in Tianshan Mountain submiages.Expermi ental evaluationswere conducted based on the new thought line and measurements among five aggregation methods. TheTM miageswere used as reference data. Expermi entalResults show that the Bicubic and Bilinearmethod presented more wonder fulper formance than the others. The validity of the two new measurements, which based on the new thought line, was also checked through comparing the effects of the different aggregation methods. Finally, the future research aboutusing the new evaluationmeasurements is also proposed.

本文暂时没有被引用!

欢迎关注学报微信

遥感学报交流群 分享按钮