Greedy Algorithm A Results
Number of matching criteria evaluations during coding with Greedy Algorithm A as the displacement increased.
All three sequences were coded using Greedy Algorithm A. Although the number of matching criteria evaluations exhibited no obvious step like behaviour, it appeared to be largely independent of the sequence coded. Despite the algorithms relatively unstructured approach to block matching the behaviour of the MSE values as the displacement increased was predictable. For the Football and Tennis sequences the MSE fell, as the displacement was increased, to a minimum at ±8 and ±6 pixels respectively, after which it began to rise.
MSE of the three sequences coded using Greedy Algorithm A with various maximum displacements.
The MSE for the Garden sequence rose immediately from a minimum at ±2 pixels and this suggests that the operative range of this algorithm was too low for it to successfully code the Garden sequence. It is also likely that the content on the Garden sequences frames led the Greedy Algorithm astray, since unlike the other Block Matching Algorithms the Greedy Algorithm was much freer to pursue local minima.
[Return to Greedy Algorithm results
[Greedy Algorithm B results
© Colin E. Manning 1996