How to compress wavelet transform

when i preform a FFT2 (of an image) it is simple to compress it by ignoring the edges and then to reconstruct using ifft.

but how can i do the same using wavelet transform?

when i preform a wavelet transform of a 2D image i get 4 different matrices (approximation, vertical, horizontal and diagonal details).

and by using:

[C,S]=wavedec2(I,level,wname);

simply provides me with a long vector C of the matrices above.

i found this:

[I_c]=wdencmp('lvd',C,S,wname,level,thr,sorh);

it suppose to compress the wavelet transform up to a cetrain threshold 'thr', but i can't figure out how to set the 'thr' parameter according to the ammount of coefficients that ware compressed.

is there any other way?