標題: 四分樹分割法全景對焦影像合成
A New All-in-focus Image Fusion Method Using Quadtree-based Segmentation
作者: 李殷
蔡文錦
Lee, Yin
Tsai, Wen-Jiin
多媒體工程研究所
關鍵字: 影像合成;全景對焦;image fusion;focus stacking;all-in-focus image;multi-focus image;fusion algorithm
公開日期: 2015
摘要: 全景對焦影像合成,是將從同一個場景拍攝,但對焦在不同焦距的所有來源影像,擷取每張影像清晰(有在焦距內)的部分,捨棄模糊(失焦)的區域,合成出一張整個場景都在焦距內的影像。本論文採用四分樹分割法,將來源影像分割成區塊,計算相對應區塊的空間頻率以區分該區塊的影像是否有在焦距內,將所有在焦距內的區塊拼在一起,合成出期望的結果。 在調整焦距時,鏡頭會前後移動,造成同一場景不同焦距的影像中,物體大小會有些微差異。因此除了新的分割法,本論文提出新的影像校正方法,利用光流法得出的影像位移向量調整來源影像的大小。
All-in-focus image fusion is to extract detailed information from different source images of the same scene to generate a new image containing more visual information than source images. This paper proposed a focus stacking method using image block segmentation. According to the spatial frequency of block, source images are separated into clear part and blurred part. The former one is selected to be part of output image at corresponding place. Notably, the proposed method also contains source image calibration, which means source images will be registered while processing.
URI: http://etd.lib.nctu.edu.tw/cdrfb3/record/nctu/#GT070256631
http://hdl.handle.net/11536/139573
Appears in Collections:Thesis