資源描述:
《基于梯度直方圖變換增強(qiáng)紅外圖像的細(xì)節(jié)-論文.pdf》由會(huì)員上傳分享,免費(fèi)在線閱讀,更多相關(guān)內(nèi)容在行業(yè)資料-天天文庫(kù)。
1、第22卷第7期光學(xué)精密工程Vo1.22NO.72014年7月OpticsandPrecisionEngineeringJu1.2014文章編號(hào)1004—924X(2014)071962~07基于梯度直方圖變換增強(qiáng)紅外圖像的細(xì)節(jié)趙文達(dá),續(xù)志軍,趙建,王鶴,王飛。(1.中國(guó)科學(xué)院長(zhǎng)春光學(xué)精密機(jī)械與物理研究所,吉林長(zhǎng)春130033;2.中國(guó)科學(xué)院大學(xué),北京100049;3.總裝備部工程兵技術(shù)裝備研究所,無(wú)錫214035)摘要:針對(duì)紅外圖像信噪比低、目標(biāo)邊緣和細(xì)節(jié)模糊等缺點(diǎn),提出了一種基于梯度直方圖變換增強(qiáng)紅外圖像細(xì)節(jié)的方法。通過(guò)分析紅外圖像的梯度直方圖,構(gòu)造出一個(gè)高斯函數(shù)來(lái)擴(kuò)展梯
2、度直方圖,增大圖像的梯度值。采用直方圖規(guī)定化法,實(shí)現(xiàn)圖像的梯度直方圖規(guī)定化,得到變換的梯度場(chǎng)。在從變換的梯度場(chǎng)重建增強(qiáng)的圖像時(shí),引入全變分(TV)模型來(lái)抑制噪聲。實(shí)驗(yàn)結(jié)果表明,提出算法的圖像信息熵比灰度直方圖均衡化和規(guī)定化算法顯著提高,明顯增強(qiáng)了紅外圖像的紋理細(xì)節(jié),為目標(biāo)檢測(cè)、跟蹤和識(shí)別提供了高質(zhì)量的圖像信息。提出的算法采用有限差分法迭代進(jìn)行求解和Visiualc++編程,對(duì)大小為480pixel×480pixel的圖像的處理時(shí)間約為40ms,基本達(dá)到了工程應(yīng)用對(duì)圖像處理的要求。關(guān)鍵詞:圖像處理;紅外圖像增強(qiáng);梯度直方圖變換;全變分中圖分類號(hào):TP39l文獻(xiàn)標(biāo)識(shí)碼:Adoi
3、:10.3788/{)PE.20l42207.1962EnhancementofinfraredimagedetailsbasedongradienthistogramtransformZHAOWen—da,XUZhi—junn,ZHAOJian,WANGHe,WANGFei。(1.ChangchunInstituteofOptics,F(xiàn)ineMechanicsandPhysics,ChineseAcademyofSciences,Changchun130033,China;2.UniversityofChineseAcademyofSciences,Beijing1000
4、49,China;3.TheFirstEnginecksScientificResearchInstituteoftheGeneralArmamentsDepartment,Wu,i’i,214035,China)*C0rre0gauthor,E-mail:TU巧538@ciomp.a(chǎn)c.cnAbstract:Accordingtotypicalproblemsoflowsignaltonoiseratioandfuzzyedgesoftargetdetailsinaninfraredimage,thisarticleproposesainfraredimageenhanc
5、ementalgorithmforimagedetailsbasedongradienthistogramtransform.Byanalyzingthegradienthistogramoftheinfraredimage,aGaussianfunctiontoextendthegradienthistogramwasconstructedtoincreasethegradientvalue.Byusingthehistogramspecificationmethodtorealizetheimagegradienthistogramspecification,thetr
6、ansformgradientfieldwasobtained.TheTotalVariation(TV)modelwasintroducedtocontrolthenoisewhilereconstructingthegradientfield.Theexperimentresultsindicatethattheimageinformationentropyofthealgorithmisimprovedsignificantlyascomparedwiththatofhistogramequalizationmethodandspecificationmethod.T
7、herefore,itenhancesimagedetailsandgivesqualifiedimageinformationfortargetdetection,trackingandidentification.Basedonthefinitedifference收稿日期:201307—26;修訂日期:20l3一O9—24.基金項(xiàng)目:國(guó)家自然科學(xué)基金資助項(xiàng)目(No.61137001)