A parametric inspired algorithm for motion estimation in video sequence

Dissertant

al-Zoainy, Salamah Athab Resan

University

University of Technology

Faculty

-

Department

Computer Sciences Department

University Country

Iraq

Degree

Master

Degree Date

2008

English Abstract

Motion Estimation in video sequences is an active topic because of its importance in many applications such as object tracking, visual surveillance systems, and dynamic 3-D scene analysis.

Moving object motion can be estimated according to their similarities in directions or amplitudes.

Three categories of motion estimation techniques are used; they are Hierarchical Block Matching, Phase Correlation and block matching algorithms (BMA).

In BMA each frame is divided into square pixel blocks, tor each block in the current frame search pixels of block in the previous frame Is the closest to the current block based on a standard matching, the relative positions define a motion vector associated with the current block.

This thesis attempts to implement accurate and faster algorithm for motion estimation to recover moving objects.

The proposed algorithm Multi-Pattern Search Algorithm (MPSA) uses two search patterns to detect objects movement depending on motion model, changing pattern dose not affected on the number of pixels listed in each step.

Motion vector obtained from MPSA is a combination the motion of objects in a 2-D scene and of 3-D camera motion, MPSA separates frame into moving objects and static background.

Camera parameters) can be recovered from the pixel coordinates of background points.

From camera movement one can find velocity and direction of moving camera.

Finally, the resulting of MPSA is compared with two techniques FS, TSS The experimental result of multi-pattern search algorithm (MPSA) tested on five video take by camera property (Sony D cam, with 990 xx power resolution, wide angle with focal length).

Algorithm execute in simple hardware has this property (1, 8 CPU, 512 RAM) and implemented by VB.

Language, also using too! (SAV Ulead Video Studio) to split video into frame.

From testing result found that FS procedure process (512) pixel for each block and TSS check 25 pixel while MPSA cheek 20 pixels only.

MPS A is faster than FS by (21.89 %) and (11.54 %) than TSS compared with MPSA, and it is more accurate than FS by (33.45 %) and (13.56 %) than TSS, from experimental result Find YIQ more accurate than using RGB color in (78.61), estimation camera velocity accurate in 71.42 %.

Main Subjects

Information Technology and Computer Science

Topics

• Algorithms

• APA
• MLA
• AMA

Language

English

Data Type

Arab Theses

Record ID

BIM-305070