Windows-based implementation of probabilistic video stabilization and compensation algorithm using Kalman Filter

Date of Publication

2007

Document Type

Bachelor's Thesis

Degree Name

Bachelor of Science in Electronics and Communications Engineering

Subject Categories

Engineering

College

Gokongwei College of Engineering

Department/Unit

Electronics and Communications Engineering

Thesis Adviser

Edwin Sybingco

Defense Panel Member


Elmer P. Dadios
Medi A. Nazar

Abstract/Summary

Quality of the intended motions in a video segment is being affected by external perturbations such as hand shakes, or moving platforms where the camera is mounted. This causes the unwanted jitters in a video that becomes additive to the intended motion.

This project aims to remove the unwanted motions in a video segment based on a previously developed algorithm following the general stabilization framework. The inter-frame affine parameters that describe the global motion of a video were measured and the cumulative transformation derived from this. Consequently, the intentional motion was estimated using Kalman filter. The difference between the cumulative transformation and the intended motion was taken as the unwanted motion and shall be used in image compensation. The undefined regions that resulted thereafter image compensation were filled using mosaicking. The project was implemented using MATLAB and is presented in a Graphical User Interface (GUI). The performance of the system was evaluated by computing for the Peak Signal-to-Noise Ratio (PSNR) of the output video. Results showed improvements when compared with the original video sequence. The system was also compared to other video stabilization software available in market.

Abstract Format

html

Language

English

Format

Print

Accession Number

TU14008

Shelf Location

Archives, The Learning Commons, 12F, Henry Sy Sr. Hall

Physical Description

1 v. (various foliations) : ill. (some col.) ; 28 cm.

Keywords

Video tape recorders; Kalman filtering--Data processing

This document is currently not available here.

Share

COinS