Dear experts,

I am facing with the following problem:

In a 8x8 image block, I can do DCT to get 64 frequency coefficients...

But I want just the highest frequency coefficient value, and its
position in
terms of row and column index (i, j) in this 8x8 block...

Do I also need to compute DFT/DCT to get all 64 frequency coefficients
then read off the highest one? Or is
there any faster algorithm, since I just want the highest one? 

The problem is that I don't know the position of
this highest frequency coefficient... so I cannot just compute, say,
the
coefficient at (8, 8)... maybe (6, 4) is the highest for some image,
or (5,8) is the highest for some other images

Can anybody give me any hint on this?

Thanks a lot,

-Walala

You can manually convolve/compute the highest frequency component, but
you may as well use a FFT routine to do that with less code.  I have a
feeling the FFT routines are optimized enough that they won't
overcompute, and will do it faster, anyway.

Highest frequency corresponds to alternating light/dark pixels.  any
higher frequencies just represent aliasing -- they don't really exist.

so take each row/column of the image, add the even pixels, then add
the odd pixels, subtract the two sums, and divide by 2pi*N (or maybe
that's 4pi*N or maybe 2pi*N*j just use your imagination).

Hi, shsh-ga,

The problem is that I don't know in advance which coefficient is my
highest frequency coefficient... out of the 64 possible positions...

The one computes even-odd activity corresponds to which coefficent?

thanks a lot for your comment,

-Walala

The highest non-aliasing frequency is twice the pitch of the pixels. 
So, for 72dpi photo, maximum spatial frequency is 36 lines/inch. 
Hence the odd/even part.  You are essentially autocorrelating a raster
line of the image with an alternating +1/-1 signal.

Now that I think about it more, are you looking for a 2D spatial
frequency, or a  1D frequency from a 2d image.  I don't have much 2D
knowledge, but I think for 1d, yu could easily put a number on the max
horizontal spatial frequency, or on the max 2d spatial frequency.

i was too sleepy when i wrote that --
"max 1d horizontal frequency, or max 1d vertical frequency"