MATLAB: How to decide upon a threshold

bestcalculatingoptimizationthresholdingWavelet Toolboxwthres

I am using the function wthresh to carry out soft thresholding on my wavelet coefficients however, I was wondering if someone could advise me and how to decide upon the best threshold to use. How is the best threshold usually determined?

many thanks in advance Bran

Best Answer

try my interactive thresholding app: http://www.mathworks.com/matlabcentral/fileexchange/?term=authorid%3A31862. Once you find a good threshld visually, try to figure out an algorithm to determine that from the histogram. As you know there is not one threshold or one algorithm that works best for all images ever taken in the universe. So you have to figure out the algorithm that works best for you.

Related Solutions

MATLAB: How dwt and wavedec are different ? here why cA1 and C have diffrent values

As you see, dwt() gives only a level 1 DWT, while wavedec gives you multiple levels. But you are correct that wavedec() simply iterates on the approximation coefficients of the previous level as expected.

Why do you say they are different? I do not see it. The first L(1) elements of wavedec should equal cA10 above

chan1 = randn(1024,1);
[cA1,cD1] = dwt(chan1,'db4');
[cA2,cD2] = dwt(cA1,'db4');
[cA3,cD3] = dwt(cA2,'db4');
[cA4,cD4] = dwt(cA3,'db4');
[cA5,cD5] = dwt(cA4,'db4');
[cA6,cD6] = dwt(cA5,'db4');
[cA7,cD7] = dwt(cA6,'db4');
[cA8,cD8] = dwt(cA7,'db4');
[cA9,cD9] = dwt(cA8,'db4');
[cA10,cD10] = dwt(cA9,'db4');
[C,L] = wavedec(chan1,10,'db4');

Now compare:

   C(1:L(1))

with

   cA10

Likewise, compare

C(L(1)+1:L(1)+L(2))

with

cD10

MATLAB: Wavelet decomposition 2D help with software

In terms of soft vs. hard thresholding. Only hard thresholding would actually buy you something in terms of compression because you would end up with a large number of zeros. For soft thresholding, you shrink wavelet coefficients toward zero but you don't get rid of them, so ultimately you don't save yourself memory that way.

In terms of which wavelets to use, for image processing generally the biorthgonal wavelets 'bior#.#' are favored in practice for image processing.

Best Answer

Related Solutions

MATLAB: How dwt and wavedec are different ? here why cA1 and C have diffrent values

MATLAB: Wavelet decomposition 2D help with software

Related Question