Date of Award

Spring 1-1-2012

Document Type


Degree Name

Doctor of Philosophy (PhD)


Computer Science

First Advisor

Jane Mulligan

Second Advisor

Henry M. Tufo

Third Advisor

Roger A.(Buzz) King

Fourth Advisor

Nikolaus J. Correll

Fifth Advisor

Min-Hyung Choi


Digital camera and smartphone technologies have made high quality images and video pervasive and abundant. Combining or stitching collections of images from a variety of viewpoints into an extended panoramic image is a common and popular function for such devices. Extending this functionality to video however, poses many new challenges due to the demand for both spatial and temporal continuity. Multi-view video stitching (also called panoramic video stitching) is an emerging, common research area in computer vision, image/video processing and computer graphics and has wide applications in virtual reality, virtual tourism, surveillance, and human computer interaction. In this thesis, I will explore the technical and practical problems in the complete process of stitching a high-resolution multiview video into a high-resolution panoramic video. The challenges addressed include video stabilization, efficient multi-view video alignment and panoramic video stitching, color correction, and blurred frame detection and repair.

Specifically, I propose a continuity aware Kalman filtering scheme for rotation angles for video stabilization and jitter removal. For efficient stitching of long, high-resolution panoramic videos, I propose constrained and multigrid SIFT matching schemes, concatenated image projection and warping and min-space feathering. These three approaches together can greatly reduce the computational time and memory requirement in panoramic video stitching, which makes it feasible to stitch high-resolution (e.g., 1920x1080 pixels) and long panoramic video sequences using standard workstations.

Color correction is the emphasis of my research. On this topic I first performed a systematic survey and performance evaluation of nine state of the art color correction approaches in the context of two-view image stitching. My evaluation work not only gives useful insights and conclusions about the relative performance of these approaches, but also points out the remaining challenges and possible directions for future color correction research. Based on the conclusions from this evaluation work, I proposed a hybrid and scalable color correction approach for general n-view image stitching, and designed a two-view video color correction approach for panoramic video stitching.

For blurred frame detection and repair, I have completed preliminary work on image partial blur detection and classification, in which I proposed a SVM-based blur block classifier using improved and new local blur features. Then, based on partial blur classification results, I designed a statistical thresholding scheme for blurred frame identification. For the detected blurred frames, I repaired them using polynomial data fitting from neighboring unblurred frames.

Many of the techniques and ideas in this thesis are novel and general solutions to the technical or practical problems in panoramic video stitching. At the end of this thesis, I conclude the contributions made by this thesis to the research and popularization of panoramic video stitching, and describe those open research issues.