Raj M. Kumar

Publication Date

1994

Document Type

Dissertation/Thesis

First Advisor

Bow, Sing-Tze, 1924-

Degree Name

M.S. (Master of Science)

Legacy Department

Department of Electrical Engineering

LCSH

Image compression--Digital techniques

Abstract

The problem mostly concerned in image communication is to increase the channel capacity, or to reduce the storage of images while maintaining an acceptable fidelity or image quality. The interframe techniques conventionally used could reduce the bandwidth significantly and were known to be able to transmit the details of only the stationary parts of the image. However, loss of resolution observed in moving areas was an objectionable artifact, which became increasingly obvious in video transmission system as the performance of cameras and display devices improved. Motion Compensation was adopted and found efficient in improving the quality of the decoded image at the expense of generating pictures of compatible quality. Nevertheless, this technique can be used to transmit signal with high spatial resolution, except those whose motion cannot be estimated accurately. The approaches in merely compressing the individual frames or reducing the image size and the frame rate are not only impractical but also insufficient for many video applications. While reducing image size or the number of frames per second could reduce the bandwidth, they will degrade the image. Therefore, Motion compensation algorithm has been introduced to supplement the still picture coding techniques, so that the image will not be degraded and in addition the amount of image information that remains constant from one frame to another has been further compressed. The estimation of motion fields from an image sequence is a difficult task for scenes which contain both moving and overlapping objects. In such cases, the problem is to simultaneously estimate the motion field and detect both the motion discontinuity and the uncovered region. In this thesis a new technique is proposed for bandwidth compression coding of sequence of digitized video images. In the method proposed two types of frame processing; namely interframe with the motion estimation and compensation, and intraframe with block truncation coding are used. The motion estimation is obtained through a blockmatching technique, with the assumption that pure translational, rotational and uniform motion within a block are involved. The regular assumption usually used on the constant intensity along the motion trajectory in the spatial-temporal path is relaxed in this thesis. That is, assuming that there is tolerable illumination changes between the reference frames and the frame to be motion compensated. Under these assumptions, the prediction error after motion compensation can be reduced, in case the translational and rigid body assumptions are violated. Motion vectors are estimated by conjugate direction search algorithm and rotational block matching. Adaptive 2-bit quantization over small blocks of space are used for the block truncation coding. Moreover with this process information about the areas either occluding, appearing or disappearing are transmitted. This process is successful in the detection of the foreground object, and correct estimation of their motion, as well as in the detection of the appearing areas, so that the block truncation coding can be applied to these particular areas. Experimental results of this research show that the approach proposed leads to the satisfactory detection of motion boundaries and uncovered regions, which, in turn leads to a more accurate motion field estimation of the motion boundaries as well as near the uncovered regions.

Comments

Includes bibliographical references (pages [123]-124).

Extent

xiii, 124 pages

Language

eng

Publisher

Northern Illinois University

Rights Statement

In Copyright

Rights Statement 2

NIU theses are protected by copyright. They may be viewed from Huskie Commons for any purpose, but reproduction or distribution in any format is prohibited without the written permission of the authors.

Media Type

Text

Download

Share

COinS