Please excuse the double post. Browser or Forum problem. Lynn
sigma10,
I have not studied or used French for many years, but it appears that the article you posted does not include any algorithm for determining the Compression Ratio. It explains how it is calculated, but it does not specify a means of selecting the thresholds for dropping or keeping bins nor does it indicate which bins are considered to be too high in frequency. It also does not report how the remaining coefficients are identified.
If I had been a reviewer on that article it would not have been published without extensive revision.
It appears that you either need to look into some of the references in that article to see if these questions have been answered elsewhere or you will need to do some of the work for yourself.
1. Define your goals: Do you want to see how much compression you can get at a specified error (or distortion) or do you want to see how much error you get for a specified compression?
2. Select a method to determine the threshold(s) which will let you meet your goal.
3. Select a method to determine the highest frequency data which will be retained.
4. Note that 3 and 4 may interact.
5. Define a storage or communication protocol which identifies the coefficients retained.
6. Run tests to verify that the system selected works as expected and that its performance is consistent with the published results.
Keep us posted on how it goes.
Lynn
