signal compression using wavelet transform

Hello, i'm doing a project using wavelets, and it cannot be using wavelet toolbox on Matlab because, first its installed at college and i do not own a copy at home, just in the lab, 2nd, it got to be on android, so just using wavelets toolbox isn't enough.

i have done pretty good so far, i've been testing it using matlab in the laboratory and my java code, with good results, now i need to compress the signal, i've been reading about it and it says something like "threshold the coefficients and set them to zero those lower than the threshold", so my question is: isn't it the same thing using for denoising?, or just by denoising i am already compressing the signal??, btw i was already able to write code for denoising.

thanks in advance and sorry for my long post.

Answer 1

[Wayne King]

Yes, it is very similar to denoising using wavelets. One difference is that with denoising you often do NOT set the coefficients to zero, that is hard thresholding, rather you "shrink" them toward zero using "soft" thresholding.

Also, even with hard thresholding, you leave the zeros in and reconstruct the signal. For compression, you get rid of the zero coefficients, thereby compressing the data, by only retaining a subset of the wavelet coefficients.

[Nelson]

thanks a lot, but now i got some questions, :

-does it have to be on both detail and approximation coefficients? or just details or coefficients??

-Does it have to be on all the detail/approximation levels??

-When, for example, my program finds that, lets suppose, being the coefficients DETAILS at level 2, my program finds that DETAILS(4)<=threshold, so i got rid of it, should i also get rid of APPROXIMATION(4), even if this is higher than the threshold, in order to keep same lengths for both DETAILS and APPROXIMATION?.

I hope you understand my questions, thanks in advance.