EE4414
Multimedia Communication Systems II
Experiment Two
Motion Estimation for Video Coding
Yao Wang and Xiaofeng Xu
1. Introduction
As you have learnt in the lecture,
motion compensated prediction is an important step in
modern video coding techniques. For this experiment,
you
are provided with
MATLAB
programs for performing motion estimation and motion compensation using different
techniques: EBMA (with integer or half-pel accuracy) and HBMA. You should go
through the MATLAB codes and try to understand how each algorithm is implemented.
You are also asked to run these programs with different parameter settings and record the
resulting prediction results and computation time. From these experiments, you should
appreciate the trade-off between prediction accuracy and computation time with different
algorithms and parameter settings. Another program applies block DCT on an image,
quantizes the DCT coefficients, and counts the number of non-zero coefficients after
quantization. The number of non-zero coefficients is a good indicator of the required bit
rate to code a frame. By applying this program to both an original video frame and the
prediction error, and comparing the resulting numbers of non-zero coefficients, you can
also appreciate the bit-rate savings by motion estimation.
This
preview
has intentionally blurred sections.
Sign up to view the full version.

This is the end of the preview.
Sign up
to
access the rest of the document.
- Information Theory, DCT, Video compression, motion estimation, non-zero coefficients
-
Click to edit the document details