how to obtain wavelet co-efficient

I am working on 3D IWT ,sir,the output i need is eight 3-D frequency sub-bands

LLLr,LLHr,LHLr,LHHr,HLLr,HLHr,HHLr,HHHr

Each level of decomposition, r, of the transform decomposes the 3-D image input into eight 3-D frequency sub-bands as mentioned,I have uploaded the block diagram ,

http://imgur.com/QLLy2

Answer 1

[Wayne King]

Hi, MATLAB supports a 3-D wavelet transform, but the 3-D lifting scheme is not supported so the coefficients will be not be integers even if you start with integer-valued data. That being said, please look at:

http://www.mathworks.com/help/toolbox/wavelet/ref/wavedec3.html

Look at the explanation for the dec field of the structure array output.

Also, look at the reference page for dwt3

http://www.mathworks.com/help/toolbox/wavelet/ref/dwt3.html

Wayne

[FIR]

sir,how can i process,please provide me some help or codingS for performing 3D-IWT,I SEARCHED LOT ,BUT COULD NOT FIND ANY

[Walter Roberson]

Then you will have to write the code yourself.

[Walter Roberson]

Wayne, the iwt2 documentation is a bit unclear, in that it talks almost strictly about 2D, but says that TrueColor (which is 3D) is supported, but doesn't talk about what the outputs would be in that situation.

FIR, is your input a 3 dimensional data set (a volume), or is it a flat image that is truecolor (RGB) ?

[Wayne King]

Oh, sorry Walter, I may have misunderstood the poster. I assumed he really meant a 3-D wavelet transform (because of the subbands he listed) which would be for a volume of data, and not a RGB image. Walter is right, if you are really interested in transforming an indexed image than the 2-D tools are right.

However, if the latter is the case, then the poster is misrepresenting the number of subbands obtained. The subbands that the poster references:

LLL,LLH,LHL,LHH,HLL,HLH,HHL,HHH

are obtained in a true 3-D wavelet transform and not the 2-D wavelet transform of an indexed (RGB) image.

[FIR]

SIR I HAVE UPLOADED MY IMAGE

http://imgur.com/TGpQK

THE SIZE IS 104 88 3

[Wayne King]

The important question is whether this MRI image is just one image in a VOLUME of data or not. An MRI dataset very rarely consists of one slice (coronal, sagittal, or axial--the case you show). If you have a dataset consisting of a volume of MRI data, say 10 axial slices at different depths of the brain, then the true 3D wavelet transform could be the appropriate tool, depending on what your goal is.

If on the other hand, this one image is an RGB image which just happens to be represented in MATLAB as 3D because of the color map.. That is a different matter, the 2D wavelet tools handle RGB images.

There is a demo, wavelet3ddemo.m, which works on a true 3D MRI volume.

[FIR]

I HAVE UPLOADED MY PAPER FOR YOUR REFERENCE

http://www.speedyshare.com/files/30552116/base_paper.pdf

[FIR]

SIR AS U SAID I HAVE MANY SLICES IN MY DATA SET,MAY BE AROUND 26 ,IS IT POSSIBLE TO PERORM,I AM CONFUSED THATS Y AM POSTING MUCH QUESTIONS ,MY PAPER IS

http://www.speedyshare.com/files/30552116/base_paper.pdf

[Wayne King]

I'm sorry I really don't have time to read your paper. If your goal is to process the whole 26 slice dataset, with an eye toward compression for example, then wavedec3, waverec3 are the right tools. Please look at the demo that I referred you to.

[FIR]

i saw that demo sir thank u am graetful to u ,

http://imgur.com/QLLy2

please look at the block diagram,and say howb to get that coefficient

Answer 2

[Wayne King]

FIR, I told you that answer. Please look at the example in wavedec3. I included the link for you above. If you open the reference page and read it, it tells you how to find those coefficients.