Transcript WebCam Presentation - Amateur Astronomers, Inc.

What is a webcam?
• Webcams are small digital video cameras that hook up
to your computer at the USB or Firewire port
• Some webcams are true CCD devices
(like the Philips ToUcam)
• The produce 320x240 pixel images
(and other resolution modes as well)
• They are lightweight & Cheap, $100 or so…
• They produce color digital video files with sound
(avi file format)
Many varieties of Webcams
• Logitech QuickCam - one of the first to be used by amateur
astronomers for Lunar and Planetary work.
• Philips ToUcam has been a very popular and inexpensive
webcam for astronomy and is the camera I use.
– an excellent entry level webcam costing around $130.
• There are higher performance (and much more expensive)
webcams used by advanced amateurs for astronomical
purposes (Luminera, DMK21F04, Point Grey).
– DMK: $390 for the camera, $199 for filter wheel, $285 for filter set.
– Lumenera: $995 for camera alone.
– Point Grey Research has some nice fire-wire minis ~$700 or so
• Avoid currently available CMOS devices, they lack
sensitivity compared to CCD based devices.
• Color vs gray-scale - filter wheels vs deBayering.
What’s Inside a typical webcam?
Lens with NIR filter
CCD chip behind window
Microphone
Video Circuit Board
USB connector and cord
How do you use it for Astronomy?
(you are going to void your warranty)
1) Remove
lens and
discard
3) Replace eyepiece with the webcam
4) Plug webcam into your laptop…
2) Add 1.25” adapter and
NIR blocking filter
If you’re lucky and persistent, you will get
some .avi video files of planets jiggling
around with fleeting glimpses of details on
the edge of visibility, a lot like what you see
through the eyepiece.
QuickTime™ and a
YUV420 codec decompressor
are needed to see this picture.
You can also extract single
frame snapshots from the
video, but they tend to be
blurry and don’t show the
detail you glimpse in the
video.
Wouldn’t it be great if we had some way to take the
information that we know is in the video, and
somehow put it all into one picture?
Well, thanks to a young Dutch amateur astronomer/computer programmer named
Cor Berrevoets, we have a FREE downloadable program named REGISTAX
which does just exactly that:
http://registax.astronomy.net/html/v4_site.html
Here’s what Registax does:
• Examines every frame of your video file
• Does a critical evaluation of its quality.
• Arranges frames in order of quality
• Lets you pick a reference frame and how
many of the best ones to keep.
• Aligns each frame with the reference frame
• Adds the frames digitally (stacking)
• This gives an enormous improvement in
signal to noise ratio (by √n).
• Uses wavelet analysis to sharpen low
contrast details in the image.
Believe it or not. This image came
from the video we saw in the
previous slide!
There are some details we need to
deal with before we start getting
pictures to rival the Hubble….
Critical
Details:
Critically
Important
Factor:
• To get good results we need to match the resolution
of the telescope to the digital sampling ability of the
webcam
• We do this by amplfying the focal length of the
telescope until the smallest resolved image details
are big enough to be realistically sampled by the
pixels of the webcam CCD
• We determine how much magnification we need
using the digital sampling theorem - also the basis
for high fidelity digital music recording and the
operation of cell phones.
The Digital Sampling Theorem
• In 1927 Harry Nyquist, an engineer at the Bell
Telephone Laboratory determined the following
principle of digital sampling:
• When sampling a signal (e.g., converting from an
analog signal to digital ), the sampling frequency
must be at least twice the highest frequency present
in the input signal if you want to reconstruct the
original perfectly from the sampled version.
• His work was later expanded by Claude Shannon
and led to modern information theory.
• For this reason the theorem is now known as the
Nyquist-Shannon Sampling Theorem
What does this all have to do with webcam astronomy?
1.
2.
The image made by the telescope optics is a two
dimensional analog signal made up of spatial waveforms
A webcam is a digital sampling device
Let’s re-state the sampling theorem in terms that relate to
telescopic imaging using a webcam:
The sampling frequency implied by the pixel spacing
on the webcam CCD must be at least twice the
highest spatial frequency present in the image to
faithfully record the information in the image.
If you violate this rule it’s called UNDERSAMPLING
Undersampling is BAD…
of Undersampling:
Aliasscanner
signalsand
- illusions,
notline
really
there
WeEffects
can illustrate
this with a digital
a radiating
pattern:
13 dpi
60 dpi
302 dpi
23 dpi
Oversampling is ok…
Undersampling is not!
How can we avoid undersampling in our imaging?
Here is a highly magnified view of a small portion of the
surface of the webcam’s CCD:
The maximum spacing of
15.8 microns on the
diagonal determines the
sampling frequency:
ns = 1 sample/15.8 m
= 1000 samples/15.8 mm
= 63.28 samples/mm
The Nyquist frequency is
exactly one half this value:
nN = ns/2
Default mode 2x2 binned, 320x240
= 32 cycles/mm
This is the maximum spatial frequency the webcam can
accurately sample in any telescope image.
Here’s how the pixels are utilized in the “higher” 640x480
resolution mode:
The maximum spacing of
11.2 microns in green pixels
determines the sampling
frequency for green light:
ns = 1 sample/11.2 m
= 1000 samples/11.2 mm
= 89.29 samples/mm
The Nyquist frequency is
exactly one half this value:
640x480 mode
No binning
nN = ns/2
= 45 cycles/mm
However, blue and red still have a Nyquist frequency of 32 cycles/mm
Here’s how the pixels are utilized in the 160x120 mode:
Pixels are “binned” into 4x4
pixel arrays with vertical
and horizontal spacings of
22.4 microns and a diagonal
spacing of 31.6 microns.
The minimum sampling
frequency is:
ns = 1 sample/31.6 m
= 31.64 samples/mm
4x4 binned
160x120 pixels
The Nyquist frequency is:
nN = ns/2 = 16 cycles/mm
Now let’s talk about the resolution of the telescope.
First, some optical definitions:
Focal length = F
Aperture = D = diameter of lens or mirror
Focal ratio = F/D
(usually written f/# as in f/8 or f/2.5 or referred to as f-number or f-stop)
image
D
F
Spatial Frequencies in the Telescope Image
Diffraction causes the image of a point source to be
spread out into a circular spot called the Airy disk:
d
d
The diameter of the disk, d, is dependant only upon the
focal ratio (f#) of the optical system and the wavelength,
l, of the light used:
d = 2.44lf# = 1.34f#
•= 1.34*f#
(for green light l=0.55m)
How do we convert this information into a spatial frequency?
d/2 = 1.22lf#
d/2
Two points of light separated by the radius of their Airy
disks can just be perceived as two points.
Raleigh Limit for Resolution
Minimum Spatial Wavelength Based on Raleigh Limit
d/2
Imagine the images of
many points of light
lined up in a row, each
separated from the
next by the radius of
their Airy disks:
The sinusoidal wave resulting from adding all the images can be used to define
the minimum spatial wavelengths present in the image lmin = d/2
The highest resolved spatial frequency, nmax = 1/ lmin = 2/d = 1/1.22lf#.
So, in the image from the telescope, we find that the maximum spatial
frequency, nmax, is given by a simple formula:
Maximum spatial frequency = nmax = 1/1.22lf#
For green light, l = 0.00055mm
At f/6, nmax = 248 cycles/mm
At f/15, nmax = 100 cycles/mm
Now that we know how to calculate this, we can “match” the maximum
spatial frequency with the Nyquist frequency, nN , of our webcam.
Setting nmax = nN and plugging it into the above formula, we have:
nN = 1/1.22lf# , which rearranges to:
f# = 1/1.22lnN = minimum focal ratio to avoid undersampling
f# = 1/(1.22* 0.00055*32) = 46
This applies to both the 320x240 mode and for red and blue images in
the 640x480 resolution mode.
An alternate expression for the maximum spatial frequency is given
by the cutoff frequency where the MTF goes to zero contrast:
Maximum spatial frequency = nmax = 1/lf#
For green light, l = 0.00055mm
At f/6, nmax = 303 cycles/mm
At f/15, nmax = 121 cycles/mm
Setting nmax = nN and plugging it into the above formula, we have:
nN = 1/lf# , which rearrange to:
f# = 1/lnN = minimum focal ratio to avoid undersampling
f# = 1/(0.00055*32) = 57
Again, this applies to both the 320x240 mode and for red and blue
images in the 640x480 resolution mode.
However, one could argue that critical sampling at the cutoff frequency
is silly, since there is no information available there.
Oversampling
• Astronomers doing high resolution solar
imaging routinely oversample by 50%
• This seems to result in higher contrast,
particularly at high spatial frequencies
Digitally sampled point image
Result of 50% Oversampling
How do we get the magnifications we need?
• Barlow Lens or Powermate
• Microscope Objective Transfer Lens
• Eyepiece Projection
A Barlow Lens is a good way
to achieve magnifications in the
range of 2x to 3x and most
amateurs already have one in
their eyepiece box.
It’s not a good idea to try to use
a Barlow lens at a significantly
higher power than its design
magnification. Spherical aberation is introduced this way and
can harm the image quality.
Stacking of two Barlows to get
4x works better.
Nagler sells Powermate image
amplifiers that work well in this
application although they are
expensive. They are available
in powers of 2x, 2.5x, 4x and
5x. They are used exactly like
a Barlow lens.
Microscope objectives are a
convenient way to gain high
magnification with excellent
optical quality.
Typically, 5x, 10x, 20x and 40x
are available. The 5x and 10x
would be useful for this purpose.
They are designed with a 160 mm
back focal length, and the front
working distance to the object
being magnified is a little less
than 160/M mm where M is the
magnification.
They are designed to work at the
stated magnification (etched on
the barrel of the lens) but can be
used at slightly higher
magnifications because we are
not using their full numerical
apertures with an f/6 beam.
The third easy way to couple a
webcam to the telescope is using
Eyepiece Projection. You need to
make a short extension tube that
fits and locks over the eye end of
the eyepiece and which accepts
the webcam adapter on the other
end.
A wide range of magnifications can
be obtained by this method which
has a long history of use for
conventional astrophotography in
the amateur community.
Magnification achieved and the
quality of the image obtained are
dependant upon the power and
quality of the eyepiece. Plössl
eyepieces and orthoscopics should
work well.
The atmosphere also affects the image:
The effect of atmospheric turbulence is to blur and bounce around the
perfect Airy disk image until it doesn’t look so pretty any more:
excellent
good
average
poor
bad
V
IV
III
II
I
}
Various
qualitative
seeing scales
A quantitative measure of seeing is the Fried Parameter, r0.
This parameter is expressed as a length and is the diameter of the largest
telescope that would be diffraction limited under prevailing conditions.
r0 varies less than 5 cm under poor seeing conditions up to values as high as 30
to 40 cm for excellent seeing at the best sites.
Here is the statistics for how r0 varied at one high altitude observatory site:
Bad Poor Average
10 5 3
2
1
Good
0.8
Excellent
0.6
0.4 seconds arc.
According to Cavadore*, the probability of getting a
good image from a single exposure is determined by the
aperture size and the Fried Parameter, r0:
By “good image” he means an exposure taken when the
wavefront error across the aperture is no more than
l/6.28 (0.16 waves).
1/P is the number of exposures you have to take to get a
single good one.
* Cyril Cavadore, Seeing and Turbulence,
http://www.astrosurf.com/cavadore/optique/turbulence/
Here is how this probability affects our chances of getting usable images with the webcam:
If we get lucky
and seeing gets
to be as good as
average, the
12.5” will give
one good image
for every 100
frames that we
take. To get 100
good frames, we
need to take
10000 frames!
In poor seeing conditions (what we have most of the time) it takes
• About 20 frames to get one good image from a 5” aperture
• More than 10,000 frames from a 12.5” aperture
• Forget about it for a 24” aperture
Stop down
when the
seeing is
unfavorable
Stopping Down
Note that when you stop down a telescope because of bad seeing,
you don’t need to use as much magnification to reach the critical
sampling focal ratios.
For example, our 24” Cassegrain telescope at Sperry has a focal
ratio of f/11.
If it were to be used at full aperture, 4 to 5x magnification would
normally be needed to reach the f/45 or so needed to avoid
undersampling with the webcam.
Putting a 10 inch off axis stop on it changes the focal ratio to f/26
so we need only a 2x Barlow to achieve critical sampling and at the
same time reduce the aperture to a more likely match to New
Jersey’s seeing.
Putting a 5 inch off axis stop on this telescope changes the focal
ratio to f/44 without any amplification, just about right for critical
sampling and a good match for poor to average seeing.
Length of Video
Planetary rotation imposes a limit on how many frames you can take
with your webcam. Emmanuele Sordini has figured this out for us
at: bloomingstars.com
Here are his recommendations for Mars, Jupiter and Saturn based on
keeping image blur smaller than the resolution of the telescope and
sampling ability of a webcam:
Noah
My night assistant
10” f/17.6 Newtonian.
Barlow lens mounted on-axis
in front of small diagonal.
Scope mounted on Losmandy
G11 Germain Equatorial.
Later installed in observatory.
Used for Mars Opposition in
2003 and high resolution
Jupiter pictures.
Moon image made with small refractor at f/6. Notice sampling artifacts.
Eratosthenes Region
Cassini Region
Mosaic of Plato Region
Taken with ToUcam coupled to 10” f/6
Newtonian with Barlow lens. EFL=176
Stopped down to 4” aperture, f/44
Seeing good to excellent.
Saturn
April. 2 2003, 2. 5x Barflow Lens 10” f/6 Newtonian
Nov. 16 2004, 5x Powermate 12.5” f/6 Newtonian
Jan. 22 2005, 5x Powermate 12.5” f/6 Newtonian
Feb. 3 2006, 5x Powermate 12.5” f/6 Newtonian
Price tag of observatory:
$4,000,000,000
$4000
Coprates (Valles Marineris)
image from Viking Orbiter
Images of Jupiter showing Oval BA
juniou
Red Spot Junior
Image taken 2/15/06 with
10” f/15 Refractor at
Sperry Observatory using
2x Barlow lens, f/30
First Light Image taken
6/18/07 with 7.25” f/14
Schupmann using 3x
Barlow lens, f/42
Seeing average.
Seeing poor to average.
That’s all, folks…
Helpful Hints
• Use a 2x Barlow for your early experiments. It gets the focus outside
the drawtube and into the webcam focal plane. You might not be able
to focus without it and you really need some amplification no matter
how lousy the seeing is. No problem if you have an SCT.
• Parfocalize an eyepiece with your webcam. Life is much easier if you
can prefocus before getting into the computer stuff. A motorized
finder is very helpful. It is amazing how much image motion you get
when you barely touch a manual focuser while you’re working at f/45.
• You must have a good finder. I doubt that even modern computerized
GO-TO scopes are accurate enough for webcam purposes when you
are using f/60 or higher. The 7-10x finder that came with your scope is
probably not powerful enough. A second finder working at 25x or
higher is a really good idea. I just attached a 3” f/10 Newtonian on the
side of my scope with a 12.5mm illuminated eyepiece giving 60x.
This works fine.
• Start with the moon. It is bright and easy to find and rewards you with
easy good results so you don’t get discouraged.
• If the seeing is poor to average, don’t waste your time with full
aperture if you have a 10 or 12” scope. Stop down to 5 to 8 inches.
Computer stuff
• The Computer
– If you are setting up outside each time, the computer probably has to be a
laptop. If you have a permanent observatory, a desktop is better.
– You need a reasonably fast windows PC (no Macs, sorry), preferrably
running 2000NT or XP. You can get by with ME, but it can be a painful
experience.
– Buy as much RAM and hard drive space as you can possibly afford, 200
gigabyte hard drive space is not too much! Get a DVD writer to use for
video file backup if you can.
– You need a free USB port to plug the webcam into.
• Webcam software: All you need is the driver and it is probably
already in your system. Look for a program called VRecord using the find
function in your computer. Run it after plugging your webcam into the USB
port. You don’t need the stuff on the CDROM that came with your webcam.
• Download Registax: It’s free!
http://registax.astronomy.net/
• Registax tutorials:
http://www.threebuttes.com/RegistaxTutorial.htm
Step by Step Procedure
•
•
•
•
•
•
•
•
•
•
Set up telescope and turn on drive.
Turn on your computer and make sure the date and time are set correctly.
Locate object in finder and insert amplifier and parfocalized eyepiece into the
focuser.
Center and focus object.
Replace eyepiece with webcam
Plug USB cable from webcam into computer.
Launch VRecord. If you have centered the object properly, you will see something
on the screen. It may even be in focus. If not, focus it.
In the options menu, make sure that the “Set Preview” box is checked and select
“Set Video Format” and verify that it is set for the default 320x240 pixel mode or
whatever mode you wish to use.
In file menu select “Set Capture File” and name the file appropriately for the
object you are observing. I like to use the name of the object followed by the date
and a serial number, such as: Mars 10 27 05 3. You don’t need to put the time
information in (or even the date) since that will be timestamped on the file when it
is saved after taking the video.
In the capture window, select “Set Time Limit”. In the dialog box be sure to check
the “Use Time Limit” box. Type in the video file length in seconds you wish to
use and say ok. I use 120 or 180 seconds for planetary work and 30 seconds for
lunar work where I plan to take a whole lot of pictures to make a terminator strip.
Step by Step Procedure, cont.
•
•
•
•
In the options menu select “Video Parameters”, and under the Image Controls tab,
turn off the full auto mode and select 15 frames per second and then click on the
tab named “Camera Controls”. Put the exposure slider to 1/50th of a second and
the gain at about 50% of full scale. If the image is too bright, select a shorter
exposure, if it is too faint select a longer one or slide the gain control to a higher
value. You also have controls for white level which may be difficult to use if
nothing in the object is supposed to be white. Auto should be off. Make small
adjustments in blue and red level until the color of the object is correct. Note: you
can make these adjustments during the daytime using a distant tree top as a subject.
The automatic controls will work in this case and you can let it do its thing, then
turn them off and save the settings for use at night. All you will have to change at
night will be the exposure or gain controls.
In the capture window, select start capture. This will give you another dialog box
which you have to click on again to actually start the capture. At the end of the
scheduled time, the file capture will stop automatically. You don’t need to save the
file since it is written to while you are taking the video.
Don’t forget to create a new capture file for your next video. If you don’t, the next
video will overwrite the last one and you will be very upset.
When you drop the file menu and select “set capture file” again. It will come up
with the name of the last file as a default. You can either type in a new name, or if
you use the serial number procedure I do, just change the last number.
Step by Step Procedure, cont.
• If your computer has enough resouces (RAM, processor speed) you will
able to run Registax on your .avi files as you accumulate them. If not,
wait until you are done observing before trying to run Registax. I have
had a number of computer crashes requiring restart because I had too
many things going on at once (I was running Windows ME at the time, it
is not a problem for later versions of Windows).
• Launch Registax. Press Select button in upper left corner of startup
screen. This will give you a file browser box which will allow you to
point to the one of your .avi files you wish to process. When you click on
the file name (even before you click the Open button) you will see the first
frame of the video in the right side of the file browser box as well as in the
Registax screen behind it. If this is the correct file, press Open.
• When the file opens, if you move the cursor over the image you will see a
square with a plus sign in the middle of it. If this square is too small (the
32 pixel one for example) processing will be fast, but the program might
get lost. I like to pick the 128 pixel box when I am in the default video
format of 320x240.
Step by Step Procedure, cont.
• Move the slider at the bottom of the screen to step through the video file.
When you find a good frame (one that is sharper, more symmetrical,
shows detail) move the cursor over the image and center the square on the
object and do one left click on the mouse.
• At this point you will see a graph with a power spectrum of the selected
frame. This is a red curve starting on the left near the top with high
intensity of low spatial frequency components and dropping off rapidly to
the right. You can also select a Quality Estimate method to use at this
point. I really like the Gradient method. Select it if it is not already
selected. You can also select the appropriate % lowest quality parameter.
I would pick 80% for a start.
• You will also see a multicolored plot of the two dimensional Fourier
transform of the image. You should see a reasonably symmetric plot with
a red region in the center. If you do, you can leave it alone and press the
button marked Align just below the blue tab also marked Align (if the red
region is too broad, use the arrows by the FFT filter box to raise the
number you see there, however, Cor’s defaults will usually work best).
Step by Step Procedure, cont.
•
When you press the Align button, you will see the image area start jiggling about
madly with the alignment box moving with it. This is a first pass at roughly
aligning all the frames, but most importantly, it is the time when the program
measures the quality of each frame by the method you selected in a previous step
(use the gradient method, it works best). When the quality estimation phase is
finished, you will see a graph with a red curve and a blue curve. The red one is a
plot of the quality of each frame as determined by the method you chose. Registax
rearrange the frame order putting the best frames to the left, the worst ones to the
right. The blue curve gives you the registration difference between each frame and
the reference frame you picked at the beginning.
•
Now you get to pick how many of the best frames to keep for further processing
and eventual stacking. Grab the slider at the bottom of the screen and move it to
the left. The vertical green line will move to the left also. As you move it the
frame number and the stack size number below slider will change. Notice that the
frame numbers appear to be random now. When you have picked the stack size
you wish to keep (the better the seeing, the more you get to keep) press the Limit
button just below the Align button.
Step by Step Procedure, cont.
• At this point you can go on and select Optimize and Stack (or just
Optimize) and the program will go through the smaller set of frames and
do a better job of aligning them up with the reference frame. A better
choice at this point, however, is to press the Create button. This takes the
fifty best frames (or whatever number you see in the box to the right of the
button) and creates a much better reference frame from them to use to do
the whole set of good frames you selected.
• Let’s say you press the Create button. When you do, the program will
rapidly optimize this small set of the best frames, stack them and give you
the chance to do some processing. When it is finished you will get a
redundant dialog box telling you to enhance image and press continue and
you have to say OK. When you do, you can play with the wavelet sliders
to perk up the image. Don’t mess around too much at this stage, otherwise
the program might not recognize the unprocessed frames as being the same
object as the tuned up reference frame. I would just move the slider under
the checked 3:2 wavelet spot to about 50% and uncheck the first two
boxes. At this point you probably are already pleased. Just wait.
Step by Step Procedure, cont.
• Press the continue button, and the redundant OK button, and then press
optimize and stack. The program now goes through the rest of the good
frames and optimizes their alignment very precisely against the created
reference frame.
• When the program has finished this process, it digitally adds all the frames
together, pixel by pixel to give the stacked image.
• You can press the preview button by each wavelet to see the contribution it
will make in the final image. Some look noisy or empty. Uncheck them.
When you do, the noise in that particular wavelet component of the image
will be eliminated from the final image. Sort of a smart low pass filter.
Some look like they have interesting detail in them. Move their sliders to
higher values to see how they can help the image. A good place to start is
with the first wavelet unchecked, the second one at 25%, the third and
fourth at 50%, the fifth at 25% and the sixth just left alone.
• Play with the wavelets until you are satisfied and then you can check the
box to hold the wavelet setting for processing the next video.
Step by Step Procedure, cont.
• Look closely at the image. You may see a red fringe on one side and a
blue one on the other. This is atmospheric dispersion and you can tune it
out! Select the RGB shift tab to the right of the screen and you will see
up/down/left/right buttons for shifting both the red and the blue
components of the RGB image around relative to the green one. To help
you do this, you can turn off any or all of the components as well. It
usually only takes a few pixels shift of the red component in one direction
and a few more of the blue component in the opposite direction to make a
significant improvement. The fringes should go away, and small scale
details may sharpen up considerably.
• Now press the Final tab near the top and center of the screen. This stage
of the process will allow you to rotate and or flip the image to correct for
odd numbers of reflections (left and right reversed, like with a star
diagonal) and make North or South up as you prefer.
• You now save the final image as either a bitmap or a JPEG 8 bit file, or
three 16 bit formats: FITS, TIFF or PNG. The FIT option is the most
accurate way to save the files as separate 16bit R, G and B components.
TIFF is like jpeg and can be compressed, however it is a 16 bit file.
• Now press Select to load another file and start all over again.
• Enjoy the cool pictures!
Ed Grafton in Houston, Texas,
imaged Mars on five successive
nights from October 16 to 21 while
the dust storm that began in Chryse
pread southward across Mare
Erythraeum and Solis Lacus. On
the 21st, the dust is the wide paleyellowish veil extending from the
center partway down. Grafton used
a 14’ SCT at f/39 with an SBIG ST402 CCD camera. North is up