2D convolution in Matlab/Octave - matlab

I = imread ("lena.jpg");
%imshow(I);
K = I;
C = conv2(I, K);
imshow(C);
I am expecting something like the following as indicated in this link.
But, my octave output is blank:
What could be the possible reason?
And, how can I obtain the expected output?

imshow() expect values between [0-255]. After your convolution all your value are way above 255. So of course when you use imshow(C), matlab/octave do a type conversion using uint8(). All your value equal 255 and the result is a white image. (0 = black, 255 = white).
You also should take into account severals things:
add the option 'same' to your convolution to preserve the original size of your image: conv2(I,K,'same')
If you only apply the convolution like that, you will obtain a strong border effect, because the central values of your image will be multiplied more time than the values in the border of your image. You should add a compensation matrix:
border_compensation = conv2(ones(size(K)),ones(size(K)),'same')
C = conv2(I,K,'same')./border_compensation
Normalize the final result (Don't take into account the point 2. if you really want the kind of output that you pointed out in your question)
C = uint8(C/max(C(:))*255)

Related

Wrong output for 2d convolution function

As the title says, the output I’m getting out of this function is incorrect. By incorrect I mean that the data are overflowing. How do I normalise the matrix correctly? Currently almost all of the pictures I get are white.
I called the function from another MATLAB file like this:
mask = [3,10,3;0,0,0;-3,-10,-3];
A = imread(“football.jpg”);
B = ConvFun(A,mask);
function [ image ] = ConvFun( img,matrix )
[rows,cols] = size(img); %// Change
%// New - Create a padded matrix that is the same class as the input
new_img = zeros(rows+2,cols+2);
new_img = cast(new_img, class(img));
%// New - Place original image in padded result
new_img(2:end-1,2:end-1) = img;
%// Also create new output image the same size as the padded result
image = zeros(size(new_img));
image = cast(image, class(img));
for i=2:1:rows+1 %// Change
for j=2:1:cols+1 %// Change
value=0;
for g=-1:1:1
for l=-1:1:1
value=value+new_img(i+g,j+l)*matrix(g+2,l+2); %// Change
end
end
image(i,j)=value;
end
end
%// Change
%// Crop the image and remove the extra border pixels
image = image(2:end-1,2:end-1);
imshow(image)
end
In a convolution, if you want the pixel value to stay in the same range
, you need to make the mask add up to 1. Just divide the mask by sum(mask(:)) after defining it. This is however, not the case you are dealing with.
Sometimes that is not the needed. For example if you are doing edge detection (like the kernel you show), you don't really care about maintaining the pixel values. In those cases, the plotting of unnormalized images is more the problem. You can always set the imshow function to auto select display range: imshow(image,[]).
Also, I hope this is homework, as this is the absolutely worst way to code convolution. FFT based convolution is about 100 times faster generally, and MATLAB has an inbuilt for it.

How can I traverse through pixels?

Suppose, I have the following image in my hand.
I have marked some pixels of the image as follows,
Now, I have obtained the pixel mask,
How can I traverse through only those pixels that are in that mask?
Given a binary mask, mask, where you want to iterate over all the true pixels in mask, you have at least two options that are both better than the double for loop example.
1) Logical indexing.
I(mask) = 255;
2) Use find.
linearIdx = find(mask);
I(linearIdx) = 255;
The original question:
How can I save only those pixels which I am interested in?
...
Question: Now, in the Step#2, I want to save those pixels in a data-structure (or, whatever) d so that I can apply another function f2(I, d, p,q,r) which does something on that image on the basis of those pixels d.
Create a binary mask
Try using a logical mask of the image to keep track of the pixels of interest.
I'll make up a random image for example here:
randImg = rand(64,64,3);
imgMask = false(size(randImg(:,:,1)));
imgMask(:,[1:4:end]) = true; % take every four columns This would be your d.
% Show what we are talking about
maskImg = zeros(size(randImg));
imgMaskForRGB = repmat(imgMask,1,1,3);
maskImg(imgMaskForRGB) = randImg(imgMaskForRGB);
figure('name','Psychadelic');
subplot(2,1,1);
imagesc(randImg);
title('Random image');
subplot(2,1,2);
imagesc(maskImg);
title('Masked pixels of interest');
Here's what it looks like:
It will be up to you to determine how to store and use the image mask (d in your case) as I am not sure how your functions are written. Hopefully this example will give you an understanding of how it can be done though.
EDIT
You added a second question since I posted:
But, now the problem is, how am I going to traverse through those pixels in K?
Vectrorization
To set all pixels to white:
randImg(imgMaskForRGB) = 255;
In my example, I accessed all of the pixels of interest at the same time with my mask in a vectorized fashion.
I translated my 2D mask into a 3D mask, in order to grab the RGB values of each pixel. That was this code:
maskImg = zeros(size(randImg));
imgMaskForRGB = repmat(imgMask,1,1,3);
Then to access all of these pixels in the image of interest, I used this call:
randImg(imgMaskForRGB)
These are your pixels of interest. If you want to divide these values in 1/2 you could do something like this:
randImg(imgMaskForRGB) = randImg(imgMaskForRGB)/2;
Loops
If you really want to traverse, one pixel at a time, you can always use a double for loop:
for r=1:size(randImg,1)
for c=1:size(randImg,2)
if(imgMask(r,c)) % traverse all the pixels
curPixel = randImg(r,c,:); % grab the ones that are flagged
end
end
end
Okay. I have solved this using the answer of #informaton,
I = imread('gray_bear.png');
J = rgb2gray(imread('marked_bear.png'));
mask = I-J;
for r=1:size(I,1)
for c=1:size(I,2)
if(mask(r,c))
I(r,c) = 255;
end
end
end
imshow(I);

How to save image while keeping dimensions and pixels same?

I have to save the image. But when I try and keep the dimensions same, pixel values change. Is there any way to keep both intact.
C=imread('download.jpg');
C=rgb2gray(C);
%convert to DCT
[r1 c1]=size(C);
CDCT=floor(dct2(C));
dct=floor(dct2(C));
[r c]= size(dCipherText);
bye=c; %lenght of message bits
for i=r1:r1
for j=c1:-1:c1-28
.....%some operation on CDCT
end
end
imshow(idct2(CDCT),[0 255])
i=idct2(CDCT);
set(gcf,'PaperUnits','inches','PaperPosition',[0 0 c1 r1])
print -djpeg fa.jpg -r1
end
Don't use print to save the picture.
Use:
imwrite(i,'downdload_dct.jpg')
print will use the paper dimensions etc defined on your figure, rather than the image data itself. imwrite uses the data in i. You don't need imshow if you just want to re-save the image.
--
Update - sorry I see now that when you mean 'scaling', you don't mean the scaling of the picture but of the pixel values, and converting back from scalars to a colour. imshow only "scales" things on the screen, not in your actual data. So you will need to do that manually / numerically. Something like this would work, assuming i is real.
% ensure i ranges from 1 to 255
i = 1 + 254*(i-min(i(:))*(max(i(:))-min(i(:))) ;
% convert indices to RGB colour values (m x n x 3 array)
i = ind2rgb(i,jet(256));
not tested!

Matlab: trouble with taking derivative of image?

So I need to take the derivative of an image in the x-direction for this assignment, with the goal of getting some form of gradient. My thought is to use the diff(command) on each row of the image and then apply a Gaussian filter. I haven't started the second part because the first is giving me trouble. In attempting to get the x-derivative I have:
origImage = imread('TightRope.png');
for h = 1:3 %%h represents color channel
for i = size(origImage,1)
newImage(i,:,h) = diff(origImage(i,:,h)); %%take derivative of row and translate to new row
end
end
The issue is somewhere along the way I get the error 'Subscripted assignment dimension mismatch.'.
Error in Untitled2 (line 14)
newImage(i,:,h) = diff(origImage(i,:,h));
Does anyone have any ideas on why that might be happening and if my approach is correct for getting the gradient/gaussian derivative?
Why not use fspecial along with imfilter instead?
figure;
I = imread('cameraman.tif');
subplot 131; imshow(I); title('original')
h = fspecial('prewitt');
derivative = imfilter(I,h','replicate'); %'
subplot 132; imshow(derivative); title('derivative')
hsize = 5;
sigma = 1;
h = fspecial('gaussian', hsize, sigma) ;
gaussian = imfilter(derivative,h','replicate'); %'
subplot 133; imshow(gaussian); title('derivative + gaussian')
The result is the following one:
If your goal is to use diff to generate the derivative rather than to create a loop, you can just tell diff to give you the derivative in the x-direction (along dimension 2):
newImage = diff(double(origImage), 1, 2);
The 1 is for the first derivative and 2 is for the derivative along the second dimension. See diff.
As #rayryeng mentions in his answer, it's important to cast the image as double.
Given a N element vector, diff returns a N-1 length vector, so the reason why you are getting an alignment mismatch is because you are trying to assign the output of diff into an incorrect number of slots. Concretely, supposing that N is the total number of columns, you are using diff on a 1 X N vector which thus returns a 1 x (N - 1) vector and you are trying to assign this output as a single row into the output image which is expected to be 1 x N. The missing element is causing the alignment mismatch. diff works by taking pairs of elements in the vector and subtracting them to produce new elements, thus the reason why there is one element missing in the final output.
If you want to get your code working, one way is to pad each row of the image or signal vector with an additional zero (for example) as input into diff. Something like this could work. Take note that I'll be converting your image to double to allow the derivative to take on negative values:
origImage = imread('...'); %// Place path to image here and read in
origImage = im2double(origImage); %// Change - Convert to double precision
newImage = zeros(size(origImage)); %// Change - Create blank new image and populate each row per channel manually
for h = 1:3 %%h represents color channel
for ii = 1:size(origImage,1) %// Change - fixed for loop iteration
newImage(ii,:,h) = diff([0 origImage(ii,:,h)]); %// Change
end
end
Take note that your for loop was incorrect since it didn't go over every row... just the last row.
When I use the onion.png image that's part of the image processing toolbox:
...and when I run this code, I get this image using imshow(newImage,[]);:
Take note that the difference filter was applied to each channel individually and I changed the intensities per channel so that the smallest value gets mapped to 0 and the largest value gets mapped to 1. How you can interpret this image is that any areasthat have a non-black colour have some non-zero differences and hence there is some activity going on in those areas and any areas that have a dark / black colour means that there is no activity going on in those areas. Take note that we applied a horizontal filter, so if you wanted to do this vertically, you'd simply repeat the behaviour but apply this column-wise instead of row-wise as you did above.

bwlabel on enhanced image results in black image

I tried to enhance an image and perform connected component analysis but it returns a black image.
My code is
I = imread('Sub.png');
I=rgb2gray(I);
imshow(I)
J = adapthisteq(I);
imshow(J)
figure, imhist(J,64)
% I = contrast(I);
L = bwlabel(J);
figure,imshow(label2rgb(L,'jet','k','shuffle'));
Also how to number each blob after bwlabel
I think that is only a matter of scaling the intensity of J when you call bwlabel, since the image is of type uint8. Its maximal possible value is thus 255.
Using this line instead:
L = bwlabel(J/255);
Outputs the following:
Yay!