Resizing and repeating an image
I am having a problem with images. I am trying to take an image and render it 1 / 4th from its original size and then repeat it as a 2x2 matrix. View like this:
Given an input image:
---------------------
| |
| |
| IMG |
| |
| |
---------------------
Firstly, shrink it to 1/4 of its original size:
-----------
| img |
| |
|---------|
Then concatenate it into a "2x2 array" of the shrunken image:
---------------------
| img | img |
| | |
|---------|---------|
| img | img |
| | |
---------------------
What I'm struggling with is the fact that I don't know how to turn it into a 2x2 array. Any suggestions? This time it is not really HW :) This is for educational purposes. Here's what I've tried so far:
function[newImg] = immultiply(picture)
Image = imread(picture); %// Reads in the picture
[r, c, l] = size(Image); %// Finds the images rows, columns and layers
rows = round(r ./ 2); %// Divides up the rows
columns = round(c ./ 2); %// Divides up the columns
newImg = cat(3,rows,columns,3); %// Creates my image, but just gives me a blank thing
imshow(newImg)
end
I will update as I work further. Thank!
source to share
The previous two answers are correct and I originally worked on it just by playing around and not going to post it, but with minor changes I think it touches on some of Luis Mendo's comments about Cammal's answer.
My original thought was, why throw data away? If you downsample to 1/4 the size but built 4 of them, you have room for all the data:
img1 = imread('myimage.png');
subplot(1,2,1),imshow(img1)
img2 = img1([1:2:end, 2:2:end], [1:2:end, 2:2:end]);
subplot(1,2,2),imshow(img2);
The resulting top-left quadrant img2
will be exactly what Kamtal's answer gives: interpolating the nearest neighbor of pixels with odd x and y coordinates. The other 3 will be (even / odd), (odd / even), (even / even). Each pixel in img1
appears in img2
, and each sub-image may be slightly different.
If instead we want to combine data from each of the 4 images into one image without throwing away all the data, we can change that a bit. We just take the average of 4 images. Note that img2
it's the same here as above, I'm just breaking out the calculations to make it obvious.
img1 = imread('myimage.png'); subplot(1,3,1),imshow(img1) img2a = img1(1:2:end, 1:2:end); img2b = img1(1:2:end, 2:2:end); img2c = img1(2:2:end, 1:2:end); img2d = img1(2:2:end, 2:2:end); img2 = [img2a img2b; img2c img2d]; subplot(1,3,2),imshow(img2); img3a = (img2a + img2b + img2c + img2d)/4; img3 = [img3a img3a; img3a img3a]; subplot(1,3,3),imshow(img3);
Here img3
shows four identical images, all of downsampling an original image using the filter medium.
To do this with a filter, you must use the kernel:
[0.25 0.25]
[0.25 0.25]
It just takes the average of 4 elements in the neighborhood. With the kernel start at (1,1), your downsampled interpolated image will be in odd numbered rows / columns:
img1 = imread('myimage.png');
subplot(1,2,1),imshow(img1)
h = [0.25, 0.25; 0.25, 0.25] //% define the mean filter kernel
img2a = imfilter(img1, h); //% filter before applying Kamtal solution
img2b = img2a(1:2:end, 1:2:end);
img2 = [img2b img2b; img2b img2b];
subplot(1,2,2),imshow(img2);
The resulting image should be the same as img3
above.
(By the way, average filtering with a 2x2 kernel and then sampling to 1/4 size is essentially bilinear interpolation. imresize
Uses bicubic interpolation by default, so its results will be slightly different.)
source to share
Just use the instructions imresize
andrepmat
i = imread('lena.png'); figure(1),subplot(1,2,1),imshow(i) [n,m,d] = size(i); newI = imresize(i,0.5); finalI = repmat(newI,2,2); figure(1),subplot(1,2,2),imshow(finalI);
newI is the image resized to 1/2. This is equivalent to shrinking the image to 1/4. finalI is the last image that is repeated 4 times. repmat
is a 2x2 matrix concatenation.
I highly recommend you check the documentation for these two functions: http://uk.mathworks.com/help/images/ref/imresize.html http://uk.mathworks.com/help/matlab/ref/repmat.html
source to share