Please use this identifier to cite or link to this item:

Title: 一個強健的不同曝光時間影像序列之移動物體偵測法
A Robust Algorithm for Moving Objects Detection in Differently Exposed Image Sequences
Authors: 吳俊霖
Wu, Jiunn-Lin
Jia, Je-lung
Keywords: 高動態範圍影像;移動物體偵測;臨界值法;型態學
High dynamic range image;Moving objects detection;Thresholding;Morphology
Date: 2006
Issue Date: 2017-12-18 17:44:57 (UTC+8)
Abstract: 隨著數位像機的普及,現在喜歡拍照的人越來越多,但是相對的,衍生出來的問題也越來越多。其中比較嚴重的問題就是相機所拍得的照片卻無法真實呈現我們肉眼所看到的景色。而其主要原因是因為一般數位相片的『動態範圍』相當的有限,而無法反應出外界景物的真實亮度。因此有學者提出了利用多張以不同曝光時間拍攝的照片來合成『高動態範圍(high dynamic range)』影像』。如此一來,卻又衍生出另一個問題,在現實的情況中,就算是同一個場景,樹葉有可能因風吹而擺動,可能有小蟲子飛過,甚至是有人物的走動,這些移動物體都會造成所合成之HDR影像有鬼影的產生。本研究的目的就是希望能夠在不同曝光時間拍攝的影像序列中,自動地偵測出這些移動物體(moving objects),並在合成高動態範圍影像的過程中把其移除,而得到一張具有高動態範圍且沒有鬼影或殘影的影像。由於我們的來源影像是數張以不同曝光時間拍攝的影像,也就是說彼此之間會有相當大的亮度(illumination)變化,所以可以想見傳統的移動物體偵測法並不能適用。本研究中我們提出,先把不同曝光時間的原始影像,利用相機響應函數(camera response curve)轉到HDR值域,即外界真實亮度(radiance map)下,接著我們即可使用一簡單的臨界法(thresholding)來偵測移動物體。由於偵測結果可能會有雜訊或在移動物體內部會有小破洞的產生,我們將利用型態學中的侵蝕(erosion)與膨脹(dilation)運算來有效解決此問題。最後我們便可以在合成HDR影像的過程中,把所偵測出之移動物體去除,進而得到一張完美之高動態範圍影像。實驗結果顯示,所提方法能在不同曝光時間影像序列中,有效地自動偵測出移動物體,並在合成HDR影像的過程中把其去除。
As digital cameras become more and more popular recently, it is very easy for us to take many digital photos. However, they are rarely true measurements of relative radiance in the scene due to the limited dynamic range in the image acquisition devices. For solving this problem, some researchers proposed methods to recover the single high dynamic range radiance map from multiple images with different exposure time. In practice, it leads another problem that the composed HDR image will appear blurry and ghosted if there are moving objects between different views of a dynamic scene.Conventional motion detection approaches can't be applied due to there are obvious illumination changes in the source image sequence. In this paper, we propose a robust algorithm to detect the moving objects in the differently exposed image sequences. We first convert the original LDR photos to the HDR domain that can be treated as the real radiance in the scene, then we can use a simple thresholding method to detect the moving objects. For reducing the noise caused in previous step and filling out the small broken holes in the body of detected moving objects, we use the morphology approach to refine the result of motion detection. Finally we are able to obtain a good HDR image by eliminating the moving objects at the HDR image composition stage. Experimental results demonstrate the effectiveness and robustness of the proposed method.
Relation: TANET 2006 台灣網際網路研討會論文集
Data Type: conference
Appears in Collections:[TANET 台灣網際網路研討會] 會議論文

Files in This Item:

File SizeFormat
688.pdf1744KbAdobe PDF130View/Open

All items in 學術集成 are protected by copyright, with all rights reserved.

社群 sharing