Section 2: Handling Images in Matlab

Matlab has some nice built-in features that make it convenient to experiment with images. This part of the assignment will introduce you to several of these features.

Where are the images?

A repository of images is growing in the directory /cs/cs153/Images and its subdirectories. For this assignment, there is a set of test images in /cs/cs153/Images/spamImages2008.

Matlab's imtool

Matlab has a nice image interface (through its image processing toolbox, which HMC's versions of matlab do have) named imtool. Simply type imtool at the matlab prompt to open a window. You should be able to open the image at /cs/cs153/Images/a1/testimage.jpg in the customary way, i.e., "File" ... "Open," etc.

You'll see the various options for manipulating and gathering information about an image -- try them all out! A wonderful one is "File" -- "Export to Workspace." When you choose that, matlab will give you a dialog box in which to specify the name of the variable that will hold the pixels of the image. For the moment, simply use A.

Back in Matlab

If you enter whos, you will now see a variable named A It is a "3-dimensional matrix" (three 2d matrices, one for each of the red, green, and blue color bands) that holds all of the pixels of the image. For example, the axes demonstrate that the upper-left corner of the image is pixel (1,1). (Matlab starts counting at 1.) That pixel has three components: a red value, a green value, and a blue value. For pixel (197,290), those components are stored in A(197,290,1), A(197,290,2), and A(197,290,3), respectively. You can check them out together by typing 
> A(197,290,:)
where the colon is used similar to the slice notation in Python. As a result, you can define a subimage: B = A(274:342, 160:240, :) and then use the imtool to show that: choose "File" - "Import from Workspace" to do so.

Warning: note that the rows are the first component of an image matrix, the columns and the second component, and the color bands are the third component.

Changing Images

You can also change the image pixels and then review the image. For example, running this code (in an m file is easiest, though not necessary) 

for row = 274:1:342,   % low:step:high
  for col = 160:1:240,
    A(row,col,:) = [0 255 0];  % red green blue
  end
end
will change the spam in A to a rectangle of green pixels.

In fact, you can take advantage of matlab's support for indexing into matrices. For example, you can accomplish the same change to the image (green pixels on spam) with the following three lines: 

A(274:342,160:240,1) = 0;        
A(274:342,160:240,2) = 255;        
A(274:342,160:240,3) = 0;
This takes advantage of the fact that when you assign a scalar to a vector or matrix, matlab automatically places that scalar into each component of the destination array. Unfortunately, the following code will not work (matlab isn't quite smart enough to figure out what you're trying to do): 
A(274:342,160:240,3) = [0 255 0];
but this one does work: 
A(274:342,160:240,3) = 255;
Reload the image A to see the difference.

More image processing

Color Bands
Pixels are the starting point for most (but not all) image interpretation. To get comfortable at fiddling around with pixels, try setting all of the green components of the image A to zero to see the effect it will have. (Remember, matlab is good at manipulating matrices. To set all of the green components to 0, you don't need to write a pair of nested for loops, you can just write A(:,:,2) = 0. Try zeroing and saturating (setting to 255) other color bands. Remember you can always reset A by re-reading the image from file.

Colorspace conversion
Matlab has a built-in rgb2hsv function. Quoting from Matlab's help system: 

HSV = RGB2HSV(RGB) converts the RGB image RGB (3-D array) to the
    equivalent HSV image HSV (3-D array).
Matlab capitalizes the names of functions about which it is providing help.

Try it out!    Create an image named HSV that contains A's pixels in hsv color space -- if you see lots of stuff printing out, you did not use the "print-suppressing" semicolon. You will be able to show that image using the imtool or matlab's alternative, imshow(H). However, the resulting image will not be the hue-only, saturation-only, or intensity-only image that is the next task for this assignment.

Once you've obtained and saved those three grayscale images, return to the spam-finding on the main assignment page .