利用可調式區塊比對並結合多圖像資訊之影像運動向量估測; Multi-frame Motion Estimation Using Adaptive Block Matching

NCUIR > College of Electrical Engineering & Computer Science > Graduate Institute of Computer Science and Information Engineering > Electronic Thesis & Dissertation > Item 987654321/8451

Please use this identifier to cite or link to this item: http://ir.lib.ncu.edu.tw/handle/987654321/8451

Title:	利用可調式區塊比對並結合多圖像資訊之影像運動向量估測;Multi-frame Motion Estimation Using Adaptive Block Matching
Authors:	王嘉銘;Chia-Ming Wang
Contributors:	資訊工程研究所
Keywords:	多圖像;移動偵測;光流偵測;可調式區塊;區塊比對;子像素;sub-pixel;motion estimation;optical flow;adaptive window;block matching;multi-frame
Date:	2000-07-18
Issue Date:	2009-09-22 11:27:19 (UTC+8)
Publisher:	國立中央大學圖書館
Abstract:	在電腦視覺的領域中，運動估測與偵查是一個有趣而重要的主題。許多的運動估測方法都以計算光流場開始，但是，我們也都了解，傳統的光流估測演算法具有雜訊而且正確性低。為了獲得更正確的光流估測，我們初步的觀念是認為盡可能利用多張影像資訊。在結合多圖像資訊的光流估測法中，我們採用以相關性為基礎的方法來實作，並且提出兩種不同的方案：一為以參考影像為主的方法，二為遞增法。在實驗當中，我們採用等速度運動的光流模組。對於光流偵測來說，求到子像素的精確度是相當有必要的。我們採用雙線性內插法及三步偵測的搜尋法。最後，我們利用可調式的區塊選擇來做樣板比對。選擇可調式區塊，是採用亮度差值累計法。在實驗結果中，我們可以清楚看到效果的改善。 Motion estimation and detection is an important and interesting topic in computer vision. Many motion estimation approaches start from the computation of optical flow. However, it is well known that the optical flow vectors estimated with the conventional methods are usually quite noisy and inaccurate. To obtain more accurate optical flow estimation of an image, our basic idea is to use as much information as possible contained in a number of image frames. In this thesis, we use the correlation-based method. Two different approaches called the reference-frame approach and the incremental approach are used in estimating multi-frame optical flow. We adopt the constant velocity flow model in all of our work. Moreover, sub-pixel refinement is necessary for every estimations. Bilinear interpolation and three-step searching are used in sub-pixel refinement. Finally, we device the adaptive window selection scheme in template matching. The adaptive windows are selected with an efficient algorithm which is called the difference accumulation algorithm. The accuracy improvement is demonstrated in our experimental results.
Appears in Collections:	[Graduate Institute of Computer Science and Information Engineering] Electronic Thesis & Dissertation

Files in This Item:

File	Size	Format

社群 sharing

Loading...