http://in.mathworks.com/matlabcentral/fileexchange/37950-feature-extraction-using-multisignal-wavelet-transform-decomposition

Try this!!!

Thank you.

The wavelet tutorial by Robi Polikar is very very helpful in understanding wavelets

http://person.hst.aau.dk/enk/ST8/wavelet_tutotial.pdf

Mat lab gives you many things in wavelet. wavelet is a frequency selective processing.

You can get the high or low frequency information of the image. based on your application we can extract low or high frequency features from the wavelet decomposition.

Dear Neha Anegondi

I  thought you mean feature extraction for images, so you need to represent images in minimum values with less error as possible. DWT can be implemented the  combination of  LPF and HPF, and feature extraction in this case is represented by LL- band.

Regards

prof. Dr. Muzhir Al-Ani

In the feature extraction using wavelet transform, we consider the varoius features(e.g. MAV, RMS etc) of the decomposition coefficient which are found by decomposing original signal in the matlab.

Instead of considering the original signal, we consider the different wavelet decomposition coefficients for extracting features.

wavelets can be used for the feature extraction by the calculation of wavelet energy in the components after the MRA.

After each level of decomposition you can find the energy in HH HL LH components and for a energy distribution based feature vector, this gives the quantification for the texture information in the input.

Best

Dr. Vinayak

Thankyou for the replies. I will try the mentioned methods.

please mention your application. What type of image you have?

Dear Neha,

                   Feature extraction using wavelet transform scheme, very much depends on the application you want to address. in basic sense you can use MATLAB inbuilt function dwt2 to get single level decomposition of an input image. this single level decomposition will give you four components viz. LL, LH, HL, and HH. all these components are in matrix form. Mostly there sizes are one fourth of the size of original image. now its up to you, to use any one or may be combination of these decomposed components for your feature selection. if you want to use a single value as a feature instead of a decomposed component complete matrix, you  can calculate mean or entropy like features from decomposed component matrix. for instance, if you select LL component as a feature generator, you can use this whole matrix as a feature or you can take its unique property like mean or entropy as a single value. 

Best of luck......