傅楸善臺灣大學：資訊工程學研究所林裕貿Lin, Yu-MaoYu-MaoLin2007-11-262018-07-052007-11-262018-07-052006http://ntur.lib.ntu.edu.tw//handle/246246/53943隨著現今數位相機以及手機相機的畫素數量大幅提昇、以及相機體積大幅的縮小，越多越多的新科技是希望能提升相片的品質。在這些技術當中，防手震技術是為了減少或是避免因為手震所產生的模糊影像。現今主要有兩類的防手震方式，分別是光學式、以及數位式防手震。 在這篇論文裡面，我們提出從四張影像合成出一張沒有手震影像的系統。利用影像合成的技術以及從實驗上的比較，證明我們能夠重建出一張亮度正常且為銳利的影像。With the boosting of number of image sensor’s pixels and the compacter working volume of today’s digital still camera or camera phone, the need for better image quality has soared and drives more newly designed image processing techniques. Image stabilization, one of these newly techniques, plays an essential role in today’s camera design. As its words suggest, image stabilization is a system designed to reduce the amount of image blur due to human’s handshake or even preventing the chances of a blur. Currently, image stabilization can be carried out by optical lens solution or by digital image processing technique. In this thesis, we propose a digital image stabilization algorithm based on an image composition technique using four source images. By using image processing techniques, we are able to reduce the amount of image blur and compose a sharper image from four source images.Chapter 1 Introduction ...................................................................................................1 1.1 Causes of blurred image...................................................................................2 1.2 Formation of blurred image .............................................................................3 1.3 Gradient Magnitudes........................................................................................5 1.3.1 Sobel Edge Detector .............................................................................6 1.3.2 Other Edge Detectors............................................................................7 Chapter 2 Optical Image Stabilization ...........................................................................8 2.1 Image Stabilization by Prisms .........................................................................8 2.2 Image Stabilization by Moving Lens.............................................................10 2.3 Image Stabilization by Moving CCD ............................................................12 Chapter 3 Digital Image Stabilization ..........................................................................13 3.1 Digital Image Stabilization by Moving Window...........................................13 3.2 Digital Image Stabilization by Higher ISO Speed.........................................14 3.3 Digital Image Stabilization on Camera Phone...............................................14 Chapter 4 Image Stabilization with Super Resolution Reconstruction ........................17 4.1 Concept ..........................................................................................................17 4.2 Feature Detection ...........................................................................................19 4.2.1 Harris Corner Detector........................................................................19 4.2.2 Analyzing Eigenvalues .......................................................................20 4.2.3 Corner Response .................................................................................21 4.3 Feature Matching ...........................................................................................24 4.4 Image Combination........................................................................................27 4.5 Experimental Results .....................................................................................28 4.6 Conclusion .....................................................................................................29 Chapter 5 Image Stabilization with Image Composition .............................................31 5.1 Feature Detection Using SIFT (Scale-Invariant Feature Transform) ............32 5.1.1 Scale-Space Extrema Detection..........................................................32 5.1.2 Keypoint Localization.........................................................................35 5.1.3 Orientation Assignment ......................................................................36 5.1.4 Keypoint Descriptor............................................................................37 5.1.5 SIFT Result .........................................................................................38 5.2 Feature Matching ...........................................................................................40 5.3 Pre-Rotation ...................................................................................................42 5.4 Binary Tree Image Composition....................................................................45 5.5 Flowchart of Our Algorithm ..........................................................................47 5.6 Experimental Results .....................................................................................48 Chapter 6 Conclusion and Future Work .......................................................................56 6.1 Dark Image Feature Detection .......................................................................56 6.2 Speed up Feature Matching ...........................................................................56 6.3 Image Composition........................................................................................567866079 bytesapplication/pdfen-US防手震影像合成Image stabilizationSIFTthesishttp://ntur.lib.ntu.edu.tw/bitstream/246246/53943/1/ntu-95-R93922085-1.pdf