计算机应用研究
計算機應用研究
계산궤응용연구
APPLICATION RESEARCH OF COMPUTERS
2014年
11期
3494-3497
,共4页
张国际%陈强%孙权森%夏德深
張國際%陳彊%孫權森%夏德深
장국제%진강%손권삼%하덕심
光照一致性校正%定量评价%Retinex%直方图均衡化%动态范围压缩
光照一緻性校正%定量評價%Retinex%直方圖均衡化%動態範圍壓縮
광조일치성교정%정량평개%Retinex%직방도균형화%동태범위압축
illumination homogeneity correction%quantitative assessment%Retinex%histogram equalization%dynamic range compression
高分辨率图像比对法是图像质量评价中简单而又行之有效的方法,其前提是两幅用于比较的高低分辨率图像具有良好的一致性,但实际情况往往很难满足这些一致性要求。为了使不能满足光照一致性的图像也能用于卫星在轨参数监测,需要进行光照一致性校正。针对如何定量评价多幅遥感图像的光照一致性问题,提出了一种定量评价方法。该方法根据Retinex原理,将图像分解为光照分量和反射分量,然后比较光照分量直方图的相似度,将这一相似度约束在0~1之间,称为光照一致性系数。实验结果表明,这种方法评价的结果与主观感受相一致,能够量化图像的光照一致性。
高分辨率圖像比對法是圖像質量評價中簡單而又行之有效的方法,其前提是兩幅用于比較的高低分辨率圖像具有良好的一緻性,但實際情況往往很難滿足這些一緻性要求。為瞭使不能滿足光照一緻性的圖像也能用于衛星在軌參數鑑測,需要進行光照一緻性校正。針對如何定量評價多幅遙感圖像的光照一緻性問題,提齣瞭一種定量評價方法。該方法根據Retinex原理,將圖像分解為光照分量和反射分量,然後比較光照分量直方圖的相似度,將這一相似度約束在0~1之間,稱為光照一緻性繫數。實驗結果錶明,這種方法評價的結果與主觀感受相一緻,能夠量化圖像的光照一緻性。
고분변솔도상비대법시도상질량평개중간단이우행지유효적방법,기전제시량폭용우비교적고저분변솔도상구유량호적일치성,단실제정황왕왕흔난만족저사일치성요구。위료사불능만족광조일치성적도상야능용우위성재궤삼수감측,수요진행광조일치성교정。침대여하정량평개다폭요감도상적광조일치성문제,제출료일충정량평개방법。해방법근거Retinex원리,장도상분해위광조분량화반사분량,연후비교광조분량직방도적상사도,장저일상사도약속재0~1지간,칭위광조일치성계수。실험결과표명,저충방법평개적결과여주관감수상일치,능구양화도상적광조일치성。
Comparison of high resolution images is a simple and effective method for image quality assessment on condition that the two high resolution images to be compared have a good state of homogeneity.However,in reality it is hard to meet the requirements.Thus illumination homogeneity correction is necessary to make the images that do not meet the demands able to satellite parameters.To quantitatively assess the illumination homogeneity of multiple remote-sensing images,the paper presen-ted a new method.According to Retinex theory,the method decomposed images into illumination and reflection.Then it compu-ted the similarity of the illumination histograms and normalized the similarity between 0 and 1 ,which was called as illumination homogeneity parameter.Experimental results demonstrate that the quantitative assessment is consistent with subjective experi-ence and can quantify the illumination homogeneity of images.