- 學位論文
以參考影像選取、細節加強、及鬼影去除作多重曝光影像融合
Multiexposure Image Fusion Using Reference Image Selection, Detail Enhancement, and Ghost Removal
摘要
一般來說，相機等數位產品所拍攝到的影像均是屬於低動態範圍(low dynamic range, LDR)影像，影像可能包含過度曝光或曝光不足的地方。為了解決此問題，我們將利用多張不同曝光度的低動態範圍影像合成一張擁有較多場景細節的高動態範圍(high dynamic range, HDR)影像。現有的高動態範圍影像技術大致上分為兩種：(1) typical HDR imaging：利用相機響應函數計算出一張高動態範圍影像，再利用色調映射將高動態範圍影像轉換為可在顯示器呈現的影像。(2) multiexposure image fusion：利用影像融合技術，給予多張不同曝光度的低動態範圍影像適合的權重再合成一張擁有較多場景細節的高動態範圍影像，故此方法在計算上是較有效率的。但是在拍攝的過程中，通常會有物體移動的狀況發生，故本研究目的即是利用multiexposure image fusion的技術並消除影像合成時，動態物體產生移動所造成的鬼影現象，合成一張擁有較多場景細節的高動態範圍影像。
並列摘要
Generally speaking, image sensors usually have a limited dynamic range or bit resolution, i.e., a single image does not provide all details in a natural scene. In fact, a low dynamic range (LDR) image always contains some over-exposed or under-exposed regions. To copy this this problem, we can capture and combine a series of LDR images with different exposures, in which each LDR image only contains some part of the dynamic range and scene details. This technique can be classified into two main types: typical HDR imaging and multiexposure image fusion. The former is using camera response function to obtain an HDR image, and then tone mapping can be used to display the HDR image into the display device. The latter is using the technique of exposure fusion to combine the LDR images to an HDR image, and this technique is efficient in computation. However, in the process of the scene which is captured by the digital camera systems. It can’t guarantee there are not moving objects in the scene. In this study, solving the ghosting problem which produced by dynamic scene and acquiring an artifact-free HDR image will be proposed.