How do I find the gradient of an image which is divided into 32*32 blocks and is having standard deviation of every particular block?

So,I have this image, of 512*512 resolution,I have divided it into 32*32 blocks and then I have found the standard deviation of every block.Now,I am required to find the gradient of the resulting image,having the same size 512*512.According to the code I need to find the gradient of resulting image I2?

This is the code I have used till now-

function stds(b)

a=imresize(b,[512 512]);

myfun = @(block_struct) uint8(std2(block_struct.data)*ones(size(block_struct.data)));

I2 = blockproc(a,[32 32],myfun);

figure;

imshow(a);

figure;

imshow(I2,[]);

Answer 1

[Massimo Zanetti]

Whatever is your image, to compute gradient use

[Gmag,Gdir] = imgradient(I);

Is that what you need?

[Prachi Sharma]

I tried this but it gives the error as below.Also I2 after the code runs is having the following dimensions-512*512*3unit8.As you can see in the code I broke it down into blocks,so it should be 16*16 double,but it is not.I am not getting this point.

 [Gmag,Gdir] = imgradient(I2);
Error using imgradient
Expected input number 1, I, to be two-dimensional.

Error in imgradient>parse_inputs (line 144) validateattributes(I,{'numeric','logical'},{'2d','nonsparse','real'}, ...

Error in imgradient (line 80) [I, Gx, Gy, method] = parse_inputs(varargin{:});

[Massimo Zanetti]

Then you can easily solve by transforming the rgb image into grayscale. Try this:

[Gmag,Gdir] = imgradient(rgb2gray(I2));

Is it ok now?

[Prachi Sharma]

I can not understand,my image gradient looks something like this.Can you tell me if it is fine or not?

[Massimo Zanetti]

I need to see also the image itself

[Prachi Sharma]

This is the image.

[Massimo Zanetti]

Are you plotting the gradient image using imshow? Try use imagesc and post the result.

[Prachi Sharma]

Attached below are the images where-

cloudtracking-is the image I am working on.

Cloudgdir-is Gdir of the image.

Cloudgmag- is the Gmag of the image.

Cloudstd-Is the image having standard deviation of the blocks of the image.

[Massimo Zanetti]

It seems that you are computing the gradient of the image where you have composed the standard deviation blocks. Of course the gradient is like that, your image is piecewise constant.

[Prachi Sharma]

As far as I understand,you mean to say that I am computing the gradient of the image after it is done computing standard deviation of the blocks in the image,right?

[Massimo Zanetti]

Yes, it is more meaningful if you compute the gradient of the original image with clouds.

[Prachi Sharma]

You mean to say after I resize the image to 512*512 ,from this point I should be computing the gradient.

[Massimo Zanetti]

yes

[Prachi Sharma]

I have changed the code to show both types of graphs.Below is the code and thank you so much for your help. :)

function stds(b)

a=imresize(b,[512 512]);

myfun = @(block_struct) uint8(std2(block_struct.data)*ones(size(block_struct.data)));

I2 = blockproc(a,[32 32],myfun);

figure;

imshow(a);

figure;

imshow(I2,[]);

[Gmag,Gdir] = imgradient(rgb2gray(I2));

figure;

imagesc(Gmag);

title('This Is The GRADIENT MAGNITUDE PLOT');

figure;

imagesc(Gdir);

title('This Is The GRADIENT DIRECTIONS PLOT');

 [mag,dir]=imgradient(rgb2gray(a));
 figure;
 imagesc(mag);
 title('GMAG OF ORIGINAL IMAGE');
 figure;
 imagesc(dir);
 title('GDIR OF ORIGINAL IMAGE');

[Massimo Zanetti]

If it helped, please accept this answer. :)

[Prachi Sharma]

Sure :)

[Image Analyst]

I don't know why you're chopping the image up into tiles in the first place. Why? Why not just work on the full resolution image?