Image processing

Seratation of colors
Segmentation
Tiny particles
Locating objects
Other image
Green whiteboard markers
Recommended problems

Our task is, to find the red whiteboard markers on the following photo.

clear all
imag1=imread('047.jpg');

This is a 3 dimensional array. It contains for each pixel the intensity of the colors in RGB order (red, green, blue)

imag1(1,1,:)

  1×1×3 uint8 array

ans(:,:,1) =

   182


ans(:,:,2) =

   169


ans(:,:,3) =

   135

The size of the image for the first two dimension is the number of pixels

size(imag1)

ans =

        3456        4608           3

It contains unassigned integers to save space

whos

  Name          Size                   Bytes  Class     Attributes

  ans           1x3                       24  double              
  imag1      3456x4608x3            47775744  uint8

We can investigate the properties of the image with imtool

imtool(imag1)

Seratation of colors

redc=imag1(:,:,1);

This contains only the information about the intensity of the red pixels. It's a two dimensional array (a matrix), can be represented as a gray image.

greenc=imag1(:,:,2);
bluec=imag1(:,:,3);

Let us plot the different color intensities!

figure
subplot(2,2,1)
imshow(redc)
title('Red')
subplot(2,2,2)
imshow(greenc)
title('Green')
subplot(2,2,3)
imshow(bluec)
title('Blue')
subplot(2,2,4)
imshow(imag1)
title('Original')

Segmentation

We create a black and white image from the gray image by having a threshold: the pixel whose intensity is higher then the threshold will be white, the rest will be black.

mask_red=im2bw(redc,0.5);

This is a 0-1 matrix, can be represented as logical matrix.

We also use the green color information, because red caps has low intensity of green pixels, while the background has high. This way we can differentiate the background from the caps.

mask_green=~im2bw(greenc,0.25);

Since they are logical we can take the element-wise and operator, tho locate the pixels where the image contains a lot of red but only a few green.

mask2=mask_red & mask_green;

subplot(2,2,1)
imshow(mask_red) % the background is also white!
title('Red')
subplot(2,2,2)
imshow(mask_green);
title('Not green')
subplot(2,2,3)
imshow(mask2);
title('Both masks')
subplot(2,2,4)
imshow(imag1)
title('Original')

Tiny particles

The function bwareaopen deletes the small particles for example this command deletes the tiny white areas, which are smaller then 30000 pixels.

mask3=bwareaopen(mask2,30000);
figure
imshow(mask3)

Warning: Image is too big to fit on screen; displaying at 25%

Locating objects

The regionprops function locates the white areas on our image, and gives back a struct which contains the information on these areas location.

caps=regionprops(mask3, 'BoundingBox');

The number of caps found:

size(caps,1)

ans =

     2

Plotting the bounding boxes:

imshow(imag1)
for i=1:size(caps,1)
    rectangle('Position',caps(i).BoundingBox,'LineWidth',2,'EdgeColor','g')
end

Warning: Image is too big to fit on screen; displaying at 25%

Other image

imag1=imread('050.jpg');
figure
imshow(imag1);
redc=imag1(:,:,1);
greenc=imag1(:,:,2);
bluec=imag1(:,:,3);

mask_red=im2bw(redc,0.5);
mask_green=~im2bw(greenc,0.25);
mask2=mask_red & mask_green;
mask3=bwareaopen(mask2,30000);

Warning: Image is too big to fit on screen; displaying at 25%

caps=regionprops(mask3, 'BoundingBox');
size(caps,1)
figure
imshow(imag1)
for i=1:size(caps,1)
    rectangle('Position',caps(i).BoundingBox,'LineWidth',2,'EdgeColor','g')
end

ans =

     2

Warning: Image is too big to fit on screen; displaying at 25%

Green whiteboard markers

imag1=imread('043.jpg');
imtool(imag1)
% Since the difference between the green and blue caps is small, we have to
% use all three colors:


redc=imag1(:,:,1);
greenc=imag1(:,:,2);
bluec=imag1(:,:,3);

mask_green= im2bw(greenc,0.45);
mask_red=~im2bw(redc,0.25);
mask_blue=  ~im2bw(bluec,0.5);

mask2=mask_red & mask_blue & mask_green;
imshow(mask2)

Warning: Image is too big to fit on screen; displaying at 25%

One of the caps is smaller, so we have to decrease the number of pixels in the definition of 'small area'

mask3=bwareaopen(mask2,10000);
imshow(mask3)

Warning: Image is too big to fit on screen; displaying at 25%

caps=regionprops(mask3, 'BoundingBox');
size(caps,1)
imshow(imag1)

for i=1:size(caps,1)
    rectangle('Position',caps(i).BoundingBox,'LineWidth',2,'EdgeColor','r')
end

ans =

     3

Warning: Image is too big to fit on screen; displaying at 25%