if, for loop, median

hi, I am new to matlab programming. I wish to write a program involving image processing, which serves the following purpose: consider a 3x3 window in a normalised 512x512 image. Compare each value of the pixel with 0.3. If the value of image pixel is less than 0.3, sort the vector, and find the median of all the pixels with value less than 0.3. And if value is more than 0.3, neglect that pixel value, and find median of all remaining pixels which have values less than 0.3. All of this should result in a new image. Thanks and Regards Manpreet Kapoor

Answer 1

[Walter Roberson]

Is this 3 x 3 window to be slid overlapping over the whole image, or is the image to be broken up in to 3 x 3 blocks? If it is to be broken up in to 3 x 3 blocks, there is a problem because 512 is not divisible by 3.

You can use sorting to find the median if you want, but there is a median() function. For example, if P are the pixels in the window, then

median(P(P<0.3))

[Manpreet Kapoor]

walter,

please dont concentrate on the size of the image, wat I want is the program to implement the logic given. I just want to see the way the functions are designed....

[Walter Roberson]

The logic for overlapping blocks is different than for non-overlapping blocks.

If you are using non-overlapping blocks, see

http://www.mathworks.com/matlabcentral/answers/5163-image-block-block-dividing

[Manpreet Kapoor]

walter,

it is overlapping window, coz the median needs to be found for every pixel window, thus forming a new image. I know the function to be used is nlfilter, wat i want to know is the sequence of commands to perform the above said function. I basically need to learn the usage of functions in a program.

[Walter Roberson]

I'm not familiar with nlfilter, but it does appear suitable

B = nlfilter(A,[3,3],@(x) median(x(x<0.3)));

The rest of the problem statement is noise, two equivalent conditions.

[Manpreet Kapoor]

hi

this command does not work, even wen I defined both A and B.... Wat to do?

[Walter Roberson]

Please be more specific about what you mean by "does not work". Do you have the image processing toolbox?

You might want to try the example given in the documentation,

http://www.mathworks.com/help/toolbox/images/ref/nlfilter.html

[Manpreet Kapoor]

walter,

I have already gone through the above document earlier, wat I am facing trouble is in the the designing of 'fun' which performs the median of the pixel values less than 0.3.

I already mentioned I am new to matlab and programming watsoever, so cant understand the implementation of 'fun'. Please help.....

[Walter Roberson]

Okay, so let's go back to you being more specific about the command not working. What exactly do you mean by that? If the cameraman example from the documentation worked for you, then my very small variation on it should work. The cameraman example had median(x(:)) and I have median(x(x<0.3)) . For an array, x(x<0.3) will return a vector suitable for input to median, and that vector will consist only of the pixels with value less than 0.3 .

[Manpreet Kapoor]

Sorry for the misunderstanding earlier, the logic u gave is absolutely correct, but the problem I have now is that nlfilter takes almost an hour to code one image, hence another command I found faster is colfilt.

can u offer any help in this....???

any help will be highly appreciated.

Thanks and Regards

[Walter Roberson]

Sorry, I am not familiar with colfilt .

I would think it likely that blkproc() or blockproc() would be much faster than nlfilter.