兩影像間平順轉換及其於環場過場之應用

Smooth View Hopping between Two Images and Its Application to Panorama Walkthrough

何佩璇(Pei-Shiuan Ho)

指導教授 ： 洪一平

共同指導教授 ： 李明穗(Ming-Sui Lee)

國立臺灣大學/電機資訊學院/資訊工程學研究所/碩士(2007年)

https://doi.org/10.6342/NTU.2007.00046

摘要

本篇論文中，我們提出一個基於影像式繪圖的架構，無需明確幾何資訊，即可產生兩張未校正影像間過渡過程的中間影像。此兩張影像分別定義為來源影像與目標影像，來源影像位於此過渡過程的起點，目標影像位於終點。為了產生在某個時間點的中間影像，首先來源與目標影像被劃分為多個區塊，區塊代表場景中能以平面趨近的表面。每個區塊各自依照自己的轉換函式變形。在本篇論文中，此轉換函式為單應性矩陣。單應性矩陣可藉由兩影像間對應的特徵點計算而得。將所有變形的區塊整合即成為中間影像。連續播放中間影像序列可產生平順的視覺轉換。更可套用縮放模糊(zoom blur)於中間影像。利用縮放模糊的特效，可使整個影像序列更引人入勝。另外我們也引入非寫實渲染(non-photorealistic rendering)來產生藝術風格的視覺轉換。

關鍵字

變形 ； 影像式繪圖 ； 影像合成 ； 視野轉換 ； 過場

並列摘要

In this work, a framework relying on image-based rendering is presented for
synthesizing intermediate views in the forward transition of two uncalibrated images
without using explicit geometry information. First, the source and destination images are partitioned into regions. Each region is a presentation of the scene surface which
can be approximated by a plane. To generate the intermediate views during the transition, each region is morphed according to its warping function individually. The warping function adopted here is derived from homography. Homography is estimated by using the corresponding feature points found by SIFT in the source and destination images. The morphed regions are combined to generate the intermediate image at a specific moment. Then the sequence of synthesized intermediate views can be played successively to produce smooth view hopping. Each morphed image can be further improved by zoom-blurring. Taking advantage of visual effect by zoom-blurring, the whole sequence of intermediate views would be more absorbing. We also introduce painting styles provided by non-photorealistic rendering, so that view hopping with artistic styles is also available in our framework.

並列關鍵字

morphing ； image-based rendering ； view synthesis ； view hopping ； walkthrough

參考文獻

[11] H. C. Huang, S.-H. Nain, Y.-P. Hung, and T. Cheng, “Disparity-based view morph-

[41] C. J. Curtis, S. E. Anderson, J. E. Seims, K. W. Fleischer, and D. Salesin,

[1] R. Szeliski and H.-Y. Shum, “Creating full view panoramic image mosaics and en-

[2] M. Brown and D. Lowe, “Recognizing panoramas.” in ICCV, 2003, pp. 1218-1225.

[5] REALVIZ Stitcher. REALVIZ.