Back to Main Page
Video Stabilization using MSER Features
Abstract
Video stabilization is an important technique in present day digital
cameras as most of the cameras are hand-held, mounted on moving platforms or
subjected to atmospheric vibrations. In this paper we propose a novel video stabilization
scheme based on estimating the camera motion using maximally stable
extremal region features. These features traditionally used in wide baseline stereo
problems were never explored for video stabilization purposes. Through our extensive
experiments we show how some properties of these region features are suitable
for the stabilization task. After estimating the global camera motion parameters
using these region features, we smooth the motion parameters using a gaussian
filter to retain the desired motion. Finally, motion compensation is carried out to
obtain a stabilized video sequence. A number of examples on real and synthetic
videos demonstrate the effectiveness of our proposed approach. We compare our
results to existing techniques and show how our proposed approach compares favorably
to them. Interframe Transformation Fidelity is used for objective evaluation
of our proposed approach.
Results
Sequence OUTDOOR
The video sequence OUTDOOR (360 frames @ 30fps CIF resolution) is captured by us in an outdoor environment. As mentioned in the paper the scene is challenging because of moving objects in the video sequence and high jittery hand motions. Our stabilization scheme using MSER features is able to robustly stabilize the video sequence as seen by the stabilized video.
Download videos [ Input Video | Stabilized Result using our method | ]
Sequence OUTDOOR_2
The video sequence OUTDOOR_2 (360 frames @ 30fps CIF resolution) is captured by us in an outdoor environment. This scene is also challenging because of moving objects in the video sequence and high jittery hand motions. Our stabilization scheme using MSER features is able to robustly stabilize the video sequence as seen by the stabilized video. This video sequence does not appear in the paper due to paper length requirements.
Download videos [ Input Video | Stabilized Result using our method | ]
Sequence LAB_SYNTHETIC
The video sequence LAB_SYNTHETIC (30 frames @10fps, CIF resolution) is captured in an indoor lab environment and does not have any object motion. We have captured this sequence using a tripod and it serves as our stable sequence which is used as ground truth. This sequence was used in the paper for objective performance evaluation.
Download videos [ Input Video ]
Sequence CROWDED
The proposed approach was tested on an challenging video sequence. We captured an outdoor video sequence CROWDED (600 frames @ 30fps, 640 X 480 resolution) in an crowded environment. This sequence has a contingent of students performing parade during a sports event at our university campus. We have introduced hand jitters with our hand-held camcorder during the video capture. The video is a challenging case due to the movement of lot of people (students). Our stabilization scheme using MSER features is able to robustly stabilize the video sequence as seen by the stabilized video.
Download videos [ Input Video | Stabilized Result using our method | ]
The videos have been put online to help fellow researchers. Kindly cite our paper in case you find our work interesting and/or use our videos for research purposes. Clarifications can be addressed to okade.manish@ece.iitkgp.ernet.in
BibTeX
@ARTICLE{,
author = {Manish Okade and Prabir Kumar Biswas},
title = {Video Stabilization using Maximally Stable Extremal Region Features},
journal = {Mult Tools and Appl},
year = {2014},
DOI = {10.1007/s11042-012-1095-z},
month = {},
volume = {68},
number = {3},
pages = {947-968}}