摘 要:針對敦煌壁畫數(shù)字圖像拼接數(shù)據(jù)量大、計算耗時長這一急待解決的技術(shù)難題,構(gòu)建了敦煌壁畫數(shù)字圖像智能拼接系統(tǒng)的總體結(jié)構(gòu);提出并實現(xiàn)了基于SURF的圖像配準(zhǔn)算法,該算法在積分圖像的基礎(chǔ)上進行快速計算,通過快速Hessian檢測子來檢測特征點;通過計算哈爾小波變換,確定其主方向及特征點描述子,并根據(jù)描述向量之間的歐式距離實現(xiàn)圖像間的特征點的匹配;最后,采用漸入漸出的圖像融合算法實現(xiàn)了圖像拼接系統(tǒng)。實驗表明:該系統(tǒng)既能滿足匹配準(zhǔn)確性的要求,又具有計算量小,計算速度快的優(yōu)點。關(guān)鍵詞:圖像拼接; 快速魯棒性特征; 積分圖像; 海森矩陣; 特征點描述子; 圖像融合
中圖分類號:TN915.73-34; TP391 文獻(xiàn)標(biāo)識碼:A 文章編號:1004-373X(2010)16-0133-04
Dunhuang Fresco Digital Image Intelligent Mosaics System Based on SURF
HE Zhao,LU Xuan-min,WANG Jun-ben
(School of Electronics and Information, Northwestern Polytechnical University, Xi’an 710072,China)
Abstract: An intelligent digital image mosaics system structure is conceived to deal with the difficult problems of large data amount and long time consuming in Dunhuang fresco digital image mosaics. A new algorithm based on SURF for image registration is presented. It is a fast computation based on the integral image through the Fast-Hessian detector, the feature points are extracted. For each feature point, the dominant orientation is assigned by computing Haar-wavelet responses, and then the descriptor is generated. Image matching is made based on Euclidean distance between these descriptors. At last, the original image and reference image are fused together by gradating in-and-out amalgamation algorithm. Experimental result shows that this system can not only meet the requirement of accuracy, but also has a small data amount and fast speed for image mosaics.
Keywords: image mosaics; speed up robust feature; integral image; Hessian matrix; feature point descriptor; image fusion
0 引 言
敦煌莫高窟現(xiàn)有洞窟735個,其中包括四萬五千多平方米壁畫和二千多身彩塑以及藏經(jīng)洞出土的五萬多件文化遺產(chǎn)。然而隨著時間的流逝,由于壁畫原料中存在鉛成分,在光照、氧氣、溫濕等外部條件作用下,色彩大多褪變,且這種現(xiàn)象仍在持續(xù)。另一方面,人類活動的影響,使敦煌壁畫不斷遭到破壞。為了更好地研究和保護敦煌莫高窟洞窟的壁畫藝術(shù),為了未來通過數(shù)字影像的方式更好地向全世界推廣敦煌文化,亟待開發(fā)一套壁畫數(shù)字圖像智能拼接系統(tǒng)來幫助解決敦煌壁畫數(shù)字圖像的自動拼接問題。
圖像拼接技術(shù)是目前的熱門研究領(lǐng)域,是虛擬現(xiàn)實、計算機視覺、計算機圖形學(xué)和圖像處理等領(lǐng)域的重要研究課題[1-2]。……