How can I take the DCT coefficients in each 8x8 block as an image?

So basically, I have to divide up an entire image into 8x8 blocks, and perform the 2D DCT for each block, outputting the results into a new image such that the new image shows DCT coefficients for each corresponding image block.

So far this is what I have;

   % code
   I = imread('lenagray.jpg');
  I = im2double(I);
  T = dctmtx(8);
  dct = @(block_struct) T * block_struct.data * T';
  B = blockproc(I,[8 8],dct);
  figure;
  imagesc(B)

just need help with making a image of the DCT coefficients

Answer 1

[Walter Roberson]

.jpg images are always RGB (3 dimensional) even when the appear to be grey, so the result of the imread() will be 3 dimensional. im2double() of something 3 dimensional results in something 3 dimensional. blockproc() of something 3 dimensional will extract 3 dimensional blocks. You then try to do a matrix multiplication of the 3 dimensional block, which is going to fail.

You may wish consider using rgb2gray()

[Osman]

but in matlab its showing me that image only a two dimensional image. THe code works to give me the DCT of each block, but i just need to an image showing each blocks DCT, do you know how to do that?

[Walter Roberson]

Do you mean that you want an graph with 64 x 64 sub-images, each showing the DCT for an 8x8 block? Remember the DCT results can be positive or negative by several hundred, and are floating point values. Perhaps you want to try imshow(B,[])

[Osman]

This is the exact questions im working on;

Write a script to divide up an entire image (assume grayscale) into 8x8 blocks, and perform the 2D DCT for each block, outputting the results into a new image (such that the new image shows DCT coefficients for each corresponding image block). [HINT]: you can use the blockproc() function in matlab to simplify this process

[Walter Roberson]

imshow() with [] is about the best you can do, but you might want to get fancier by putting together the coefficients for each of the Y, Cr, Cb panes to produce a colored version of the data.

Answer 2

[GOKUL REDDY]

I = imread('lenagray.jpg');

I = im2double(I);

T = dctmtx(8);

dct = @(block_struct) T * block_struct.data * T';

B = blockproc(I,[8 8],dct);

figure;

imagesc(B)