INTRODUCTION TO IMAGE PROCESSING

August 21, 2007

Reading: article from Sky and Telescope: Sky on a Chip: the Fabulous CCD

OPENING AN IMAGE in HOU

Opening the HOU program:  the Hands On Universe image processing program and files are accessible from any phyiscs-floor networked school computer. After logging on look forpeither a red/green HOU icon on the Desktop (double-click on it) or for the program listing under Start/All Programs.  If neither of those works, search for HOU executable file by right-clicking on Start, then Explore; navigate to the directory C:\Program Files\HOU-IP\ and double click on the red/green HOU icon (or the .exe file)

Installing HOU on your computer (not required): navigate to T:\Software\Physics\HOU\HOU_Installer_forPC\ and double-click on the setup icon

Opening an HOU image: Under File, select Open, (the directory images should come up), then double click on Images-High School, then on 2browsers_guide_to_the_universe, then on browser5

Maximizing the image space: first click the rectangle in the upper right corner (this maximizes the window on the screen); then drag the browser 5 image's blue bar up as high as it will go in the dark gray area; then click on the square in the upper right corner of the browser 5 frame

The screen should display a black-and-white image of an astronomical object. Investigate the options under the View menu (i.e., Tool Bar, StatusBar, Color Palette Bar, and Display Controls Bar; click these on and off to see what each name refers to. After experimentation, you will want to have the ToolBar, StatusBar, Color Palette Bar, Display Controls Bar "on" and the others "off."

By this time you know something about CCDs and understand the concepts of pixels and digital imaging.
Locate -- on the screen -- the three important numbers associated with each pixel: the x, y location and the Counts value.
Each pixel is 'colored' the appropriate shade of gray according to the Counts value measured for that pixel; the translation table from 'Counts' to 'color' is the palette at the right of the screen. Experiment with the zoom by increasing the zoom factor to 2, 4, 8, 16, all the while looking at the object's center. [To keep track of where the object's center is, note that it continues to have the same x-y pixel coordinates at the bottom right of the screen (as long as the Status Bar Option has been selected). Have you also noticed that the x-y coordinates and the Counts number change as you move the mouse cursor around the screen?] At what zoom factor can you first begin to see the actual pixels?

Note that you can also zoom on only part of the image by using the Zoom box command under Data Tools. The cursor symbol changes to a square of 4 squares. Click and drag the mouse to create a rectangle in the image that you would like to zoom in on. A new pop-up window appears with the zoomed area. Did you notice the zoom factor for the zoom box? Can you zoom in on the zoomed area?

to remove any red marks on the image:  click on the "broom and bucket" icon just below the menu bar.)

FALSE  COLORING

Return to Zoom 1 on the original Browser 5 image, maximize the image, and center the object in your window. The black-and-white image is not very interesting because of the lack of detail. Can you tell what is in the image?

Astronomers use the idea of false coloring to bring out contrast and detail.

Under the File Menu Bar, click on Load Color Palette.... select rain.pal
I think it's the most interesting; you can experiment with other palettes later if you want.

Under the View Menu Bar, click on Color Palette Bar so that it is displayed on the right hand part of the screen... move the mouse/cursor around on the screen over different color areas of the image.... does the color match what the Counts value and Color Palette Bar predict?

After you finish the lab, I will ask you to describe in a quantitative manner how false color digital imaging works; i.e., what does the computer actually do (given the minimum and maximum counts recorded on the image) in order to display the false-colored image?

Your goal is to fiddle (quickly and efficiently!) with the false coloring so that the object's detail is brought out; you do this by changing the min and max in the boxes under the menu bar (and on top and bottom of the color palette). Upon opening the image, the min was probably set to 215 and the max to 459). Some hints for setting the min/max follow:

a) To find a good min,

move the mouse around over the very blackest parts of the image and notice the Counts readings; select a min value that reflects these lowest Counts values

There are three ways to adjust the min:

(1) click in the box next to the word Min and then type in the desired value,

(2) drag the red bar left or right to decrease or increase the min,

(3) click on the left-pointing arrow box....

After you make a change in the min, the HOU software should immediately refresh the image to your new standards

b) To find a good max,

start with the max given; what happens to the image as you decrease the max?

what trade are you making as you decrease the max?

c) The image will probably still not look very good. Now click on the box for Log Scaling.
The result should be impressive. You want to do some quick additional fine tuning with
min and max.   But don't agonize: there is no correct answer.

At some later point, I will ask you to describe how log scaling works. For the moment,

notice (and record) what has happened to the Color Palette bar during the switch from
linear to log scaling.

From this point on, anything underlined should be answered in your lab book.

Record your 'best' values for min and max.

Can you trace the galaxy's spiral arms to the edge of the image window? How much of the image frame would you say the object occupies? Can you see the knots of bright hydrogen in the spiral arms?

More Images & Tools

• The goal of the following parts is two-fold:

a) to get a taste of image processing

b) to learn something about some astronomical objects we'll encounter later in the course

• Now open (same path as above) browser2.  Can you tell what it is?
  (I would strongly suggest going back to a grey-scale image here.)

How many atmospheric bands can you see on the planet?
Do you see any moons?  How many?
Can you find an appropriate max/min/log combination that allows you to see all moons
AND the atmospheric bands?  If so, record the values.

• Open browser3.  What's this object?       Zooming out might help.
  

A number of pixel rows on the left side of the image have been saturated (i.e., the
maximum number of electrons possible on a pixel has been reached, and electrons
have spilled over onto adjacent rows of pixels; it is the equivalent to overexposure on film.)

Measure the sun's diameter in pixels.  How did you do it?   I can't envision communicating how you did it without a labeled diagram.

What are the bright spots over the limb of the sun?  (use the sun chapter in your text)

• Open browser4. 
  Almost always, there is information about the object in the image along with where,
  when, and how the image was taken encoded in the image; to access it,
  click on Data Tools at the top, the Image Info
  
      what is the name of the object?
      exposure time?
      time and date of exposure?
      where was the image taken? (hint: look at the telescope name and the time)

What similar objects to this are in the Stellar Evolution summary? 
    (the Stellar Evolution Summary is a clickable link from the course page)
    Notice the white dwarf star at the very center.

Starting from the original image as loaded, how many times must you click on zoom
in before you can start seeing the individual pixels?

• Open the image fireball under the directory 4jupiter_crash

    The fireball in the upper left corner of the image was produced when a piece of
    Comet Shoemaker-Levy smashed into Jupiter in 1994.         
    The known diameter of Jupiter is 143,000 km. 

Find the diameter of the fireball; show your work! remember the lab guidelines
is the fireball larger than Earth?

• Open the new browser7
  Where is this object in the Stellar Evolution summary?   What is its name?

This remnant is one of the very few that contains a visible pulsar (or neutron star) at its
center.  To find out which of the central stars is the pulsar go to figure 23-4 in Universe. 
What operation(s) do you have to do to the image to make it match exactly the picture in
figure 23-4? There are a number of ways to manipulate images found under the
Manipulation menu at the top (Rotate, Translate, Flip...)

Are the rotations produced clockwise or counterclockwise?
Tell me what operations -- in what order -- you used to make the new browser image
exactly match the figure in Universe 23-4.  (There is more than one correct answer.)

• And now back to linear scaling vs log scaling:
  For each type of scaling, I want you to figure out how the program selected the dividing lines (in Counts) between the different colors. 
  
  Suppose that an image displays n colors and that Linear Scaling is used, and that the maximum Counts value is Max and the minimum counts value is Min.  In the first HOU image you loaded (the galaxy) had n = 16, max = 215, and min = 459.  How do the ranges of Counts values allotted to each color compare?  Assuming that only the active range of counts (from min to max) is allotted one of the n colors, calculate the Counts Range (CR) of each color (first do this in symbols, i.e., with n, Max, Min) and then calculate a number using the numerical values just quoted.  Then come up with a formula for each color boundary; use i to denote the number of the color boundary, with i =0 for the lowest color boundary (which will have the Counts value = Min) and with i = Max (which will have the Counts value = Max).  As before, do this first in symbols (you might first want to use i, Min, and CR; but then remember that you have a formula for CR), and then with numbers along with the symbol i. 
  
  OK, that was the easier one.  Now do the same thing for Log Scaling.  Remember that the difference between the upper and lower values of the Counts assigned to each color was the same for linear scaling.  What about the upper and lower values of the Counts assigned to each color is the same for Log Scaling.  As before, the ultimate goal is to come up with a formula for the ith color boundary in terms of i, n, Max, and Min.
  

• For the last few activities,  we'll use the advanced features of the ds9 image
  processor used to analyze Chandra X-Ray Observatory images.
  
  
• Opening the ds9 program:  ds9 in installed on all physics-floor networked school computers.  Look for either a yellow sunburst icon on the Desktop (double-click on it) or for the Program under Start/All Programs.  If neither of those works, search for ds9 executable file by right-clicking on Start, then Explore; navigate to the directory C:\Program Files\ds9\ and double click on the yellow sunburst icon (or the .exe file)
  
  

        1) Under Analysis, click on Virtual Observatory;
            a box should pop up showing several different image servers

        2) click in the Rutgers X-ray Analysis  box;  almost immediately the box should turn
            green, and a browser window opens that lists images that can be loaded

        3)  a list of clickable image links should show up; click on the Cas A image

        4)  you should get an acknowledgement that the image has been loaded;
                then go back to the ds9 image processing window to see the image

        Cas A is the remnant of a supernova explosion;  see Universe, figure 22-23

  The scale of the image is 0.5" (seconds of arc)/(Physical) pixel. 
Determine the diameter of Cas A in pixels, seconds of arc, and then light years.
  The distance to Cas A is approximately 11,000 c-yrs. 
   You will need a large, labeled diagram to accompany your equations and
   calculations of the diameter in light years.

  We believe that Cas A exploded in 1670. 
Find the average expansion speed of the outermost remnants
(in km/s and as a fraction of the speed of light?)

advanced image processing

• creating spectra (continue using the Cas A image)

    to obtain a spectrum of this object:
    under Analysis, select Chandra Ed Analysis Tools,
    then select Quick Energy Spectrum Plot
    zoom in on spectrum by creating a box using click/drag/click; 
        right click to return to previous plot; more display options under View
    use the following link to identify the element producing 2 or 3 of the emission lines:
use the link identifying lines in supernova spectra

• x-ray vs. optical views 

    load the Coma image from the Virtual Observatory;
    Coma is a cluster of galaxies 400 M light years distant; image shows intergalactic gas
    select b for Color; select sqrt for Scale
    under Color, select Contrast/Bias; select  1.1 for the Contrast and 0.4 for the Bias 
    to retrieve the corresponding optical view:
    under Frame, select Tile Frames
    under Analysis, select DSS server; click Retrieve in the pop-up box
    After the optical image loads, click on the x-ray image
    under Frame, select Match Frames > WCS
    under Frame, select Lock Crosshairs > WCS
    under Edit, select Crosshairs
    you can click&drag the crosshairs to see which positions in the two images match

what objects are you seeing in the visible?  what in the x-ray?  (see Universe fig 26-23)

• light curves & periodicities

load the Cen X-3 image from the Virtual Observatory;|
Cen X-3 is a binary system, containing a neutron star (the core remnant of a supernova)
    and a supergiant star
under Analysis, under Chandra Ed Analysis Tools, select FTOOLS/Light Curve;
click OK 
zoom in on light curve by creating a box using click/drag/click; right click to go back;
zoom until graph displays about 100 seconds of time; do you notice a periodicity in the data?
if so, estimate the (pulsar) period;
under Analysis, under Chandra Ed Analysis Tools, select FTOOLS/Power Spectrum;
click OK 
(this command does a fast Fourier Transform on to search for periodicities in the data)
find the most likely frequency (where the peak is) and then convert that frequency to a
period;  this is the rotation period of the neutron star!