Motion cues and saliency based unconstrained video segmentation

The segmentation of moving objects become challenging when the object motion is small, the shape of object changes, and there is global background motion in unconstrained videos. In this paper, we propose a fully automatic, efficient, fast and composite framework to segment the moving object on the...

Full description

Saved in:
Bibliographic Details
Published in:Multimedia tools and applications Vol. 77; no. 6; pp. 7429 - 7446
Main Authors: Ullah, Javid, Khan, Ahmad, Jaffar, Muhammad Arfan
Format: Journal Article
Language:English
Published: New York Springer US 01.03.2018
Springer Nature B.V
Subjects:
Automobile driving
Computer Communication Networks
Computer Science
Cues
Data Structures and Information Theory
Drivers licenses
Experimentation
Multimedia Information Systems
Object motion
Optical flow (image analysis)
Salience
Segmentation
Special Purpose and Application-Based Systems
Visual task performance
Saliency
Moving object
Homography
Segmentation
Optical flow
Recognition
ISSN:1380-7501, 1573-7721
Online Access:Get full text
Tags: Add Tag
No Tags, Be the first to tag this record!
Description
Summary:The segmentation of moving objects become challenging when the object motion is small, the shape of object changes, and there is global background motion in unconstrained videos. In this paper, we propose a fully automatic, efficient, fast and composite framework to segment the moving object on the basis of saliency, locality, color and motion cues. First, we propose a new saliency measure to predict the potential salient regions. In the second step, we use the RANSAC homography and optical flow to compensate the background motion and get reliable motion information, called motion cues. Furthermore, the saliency information and motion cues are combined to get the initial segmented object (seeded region). A refinement is performed to remove the unwanted noisy details and expand the seeded region to the whole object. Detailed experimentation is carried out on challenging video benchmarks to evaluate the performance of the proposed method. The results show that the proposed method is faster and performs better than state-of-the-art approaches.
Bibliography:ObjectType-Article-1
SourceType-Scholarly Journals-1
ObjectType-Feature-2
content type line 14
ISSN:1380-7501
1573-7721
DOI:10.1007/s11042-017-4655-4