Indoors, outdoors, it doesn't matter how
wide an angle of view your digital camera lens might have, in our experience there are
invariably situations when it just isn't enough. There are often
times when you find that you can't fit into a shot all that you
want. The easy answer is to move backwards, further away, but in a
lot of situations this just isn't possible. Either there just isn't
the room to do so, or what you are
trying to take a shot of is then obscured, in part or in whole, by
something that stands between. It's particularly the case if you
tend to take shots in cities, where the buildings are always big and
the space between them is invariably small. But it can happen
anywhere at any time.
If you have a digital
SLR with interchangeable lenses then of course you do have the
option of changing the lens you have for one which gives a wider
view. But even here you often come upon the same problem, that as
wide as you can go, it's still not enough. And of course, wide angle
lenses for current DSLR's are few in number, and very expensive.
Even if you do have a wide angle lens you can use, constantly
changing from one lens to another is always awkward, especially on
city streets. Just keeping a wide angle lens on, even a zoom type,
has never for us proved possible. Whilst you might want to capture
as much of a building or scene as possible, it is also the case that
telephoto and close up shots are equally valuable.
However, perhaps the biggest drawback of all is that as you use a wider view
to get in all you want horizontally, or vertically, the other aspect
also increases. The result is that almost invariably what you are
trying to capture, the main focal point of the image, becomes
smaller as a total part of the shot. Cropping the image to
exclude too much foreground or sky that is not required may give you
a panoramic view type image, but the problem that then often arises is that the image
quality is not as high as you would like, because the image has had to be increased in size more than it normally would, reducing the
overall resolution. It's not a problem
that's restricted just to digital cameras of course, users of film
cameras have suffered in just the same way.
In the past with film
this has led to the production of special panoramic cameras, made specifically to produce panoramic images.
Most of these have been designed around the use of the standard 35mm
and 120 film formats. Those using 120 film use special wide angle
lenses, whilst some using 35mm film have lenses
that are not fixed in place on
the horizontal plane but pivot about the lens nodal point as the image is shot
thus taking in a wider, panoramic view. In more recent times
Hasselblad have made a panoramic camera called the X-Pan which also
uses 35mm film to produce panoramic shots, but like the 120 format
designs this uses specially
made wide angle lenses instead.
All types of panoramic camera are
naturally very expensive, but for those that need them they do
overcome the problem of low resolution images, producing high
quality high resolution shots, especially those that use 120 format
film. However, nice as they
are for those that can afford them, panoramic cameras don't overcome
the lack of room to shoot a scene when that particular problem
raises it's head.
Stitched Images
One of the advantages of
digital imaging is that there is a way of overcoming these problems,
and this is by stitching images together on the computer using
stitching software. It's
not restricted only to those that shoot using digital, film users can
also use the technique by scanning their images in and producing
digital files that can be worked on, but there is no doubt that,
along with the right equipment and technique, it's far easier to achieve
using a digital camera with it's LCD screen for comparing taken
images, than a film camera. Not
only that but it can be, in comparison to using film, a lot
cheaper. Although you need stitching software with which to assemble
and merge the individual images used, no special camera is required
to take them,
and the most basic low resolution digital camera can be used if you
wish. Another advantage is that even in this situation a high
resolution image can result, as a stitched image is naturally
composed of multiple shots.
Generally, stitching
images together to provide a bigger view is regarded as the way to
provide panoramic images, but is is also the means whereby an image
of a scene can be produced that might otherwise be impossible. This
can be achieved by stitching multiple rows of images together. Not all
stitching software can do this however, which is why the general view
of such software has arisen that it's only for producing panoramic
shots. The ability to take and assemble multiple row images gives
the user of digital cameras and stitching software another big
advantage over the dedicated panoramic cameras, which of course can
only provide one type of fixed format image.
In actual fact three
different types of panoramic images can be produced. The flat rectilinear, sometimes referred to as a
planar. The cylindrical, which is basically a series of rectilinear
images with the ends joined together giving up to a 360° view. And
the spherical, which can give a 360° horizontal and 180° vertical
view, virtually 360° all round. If you use the best available
software you can take a series of
shots and decide later how they should be assembled, in other words
how wide the view should be, or the perspective involved, all of
which are adjustable.
Most
panoramas can be produced in Apple's Quicktime Virtual Reality format [QTVR]
and viewed using the free Quicktime player - www.apple.com/quicktime
. These virtual panoramas
can be used to give virtual tours around almost any location or the
inside of a building, room by room. Tours such as these
are becoming increasingly common on web pages. Estate agents and
those selling cottage holidays are finding them very useful for
generating custom. The advantage of a QTVR format panorama over a
short movie file produced by rotating a video camera around 360° is
that, as a series of still images, there's no camera shake or jerky
panning, it can be panned and rotated at will and stopped for
examination at any point of interest. Great for showing the
interiors of historic buildings as floor and ceiling can be viewed
as well as the walls.
All
image types can of course also be printed to produce panoramic views
of one kind or another, which is probably of most interest to the
ordinary digital camera user.
Taking shots for
stitching
Images shot with the
intention of stitching them together later need to be done with some
care if a successful outcome is to result, and a few
basic rules need to be observed. Following the rules is
fairly easy, but it helps if you understand why you need to follow
them, and what can go wrong if you don't.
The
Lens Nodal Point.
One
of the most important points to be aware of, and understand, is that
of the lens nodal point. It is around this point that ideally the
camera you are using should be pivoted when taking images. There is
a very good reason for doing this, and it's to do with perspective
alignment and the avoidance of parallax error
All
dedicated panoramic cameras take the image in one go using a fixed
focal length lens so the perspective across the frame is constant. The panoramic types that use moving lenses
differ slightly in that instead of the whole film being
exposed at once it's done more slowly in stages as the image from the lens
passes over the film from one side to the other. A constant
perspective is still maintained by ensuring the rotation of the lens
is around it's
nodal point. Physics tells us that
the nodal point of a lens is situated along the lens axis where the refracted lens image light paths
intersect, fig 1, the image a lens produces being inverted, and this varies according
to focal length.
nb. some people refer to this point as the exit
pupil point, no doubt because the human eye works in the same basic
manner.
Lens
Nodal Point
When
taking multiple images to stitch together, the procedure is basically the same as that which the panoramic cameras
with rotating
lenses use, except it's being done in a series of separate stages,
individual images, rather than in one go.
There is therefore still the need to rotate the lens used around the nodal
point of the focal length used to ensure that the elements in adjacent images stay constant
in their perspective relationship to each other. If this isn't done,
and the parallax errors that result are too great for the stitching software
to overcome, a poor stitch results. Image elements may be out of
alignment, i.e. not match up, or 'ghosting' may result, a shadowy
secondary image element behind the first, slightly out of alignment.
To
give an example, take two poles, one in front of the other. Looking
straight on you should see the first, but not the second, which is
hidden by the first. If you turn to your left or right, pivoting
about your centre you should still not be able to see the second
pole. The same happens when a camera is pivoted about the lens nodal
point. Only if you move sideways left or
right will the second pole become visible, because you have changed position and the
perspective has changed in relation to the two poles.
This will also
happen if a camera is not pivoted about the lens nodal point. It's
view perspective changes. These perspective
changes, called parallax error, mean that in adjacent shots distances between
objects in a scene do not remain constant but alter, and the
stitching software then has difficulty registering the images
because of it.
The further away
from the nodal point that the camera is pivoted between shots, the
greater the parallax errors, and the harder
it will be for the stitching software to merge the images successfully
and obtain the correct perspective.
The most critical
situation occurs where the focused distance is minimal, or there is
close foreground detail. The greater the focused distance the less
critical this becomes. As the nodal point varies with the focal length of the
lens finding it is not a case of mathematics, but a practical matter of trial and error
adjustments.
Finding
the Lens Nodal Point.
As
the nodal point of any lens depends on it's focal length, the way of
finding it is by moving the camera forwards and backwards until a
pivot point is found where little or no parallax error can be
observed. It's done purely by subjective sight assessment, and it's
impossible to do without the use of equipment, a panoramic head,
which we will deal with later.
It
is particularly noticeable that trying to pivot the camera around a
point behind the nodal point range, generally where camera tripod
mounts are situated, produces much greater parallax errors than
pivoting around any point in front of the camera body.
Overlapping
Images
All
the shots that are going to comprise a stitch must
overlap each other by at least 30%, whether it's a single row or
multiple row stitch. Overlapping of shots is
vital as no stitching
software is able to butt join images, the
software requires comparable image elements - pixels - on which to base
the calculations that are used to manipulate the images for
merging together and warping to the correct perspective. It's better
to allow more overlap than less, anywhere between 30%-50% should
ensure a decent match, see fig 2. This figure applies whether taking
shots horizontally or vertically i.e. in normal or portrait orientation.
Alignment
Miss-alignment
of shots, vertically, horizontally or around the lens axis can lead
to difficulties not only with matching images but also to the stitched
results where excessive cropping is required, and some of the
required scene elements end up missing.
Warping
All images in a sequence must be shot from the same viewpoint
using the lens at the same focal length setting. This enables
the software to 'warp' the images. The warping of images is the means whereby the
perspective of the images is aligned to match each other. Most
decent stitching software now uses the EXIF data from the images
used to determine the lens focal length used and the resulting
perspective, correctly applying the relevant LMF factor involved.
It's important to note that different view perspectives are involved
depending on whether the images used are horizontal or vertical in
nature.
Exposure
The final
requirement is exposure. If you are taking a series of overlapping
images covering a wide perspective where the light levels might well
differ, it's important to use a constant exposure setting to
maintain the same level of exposure across the series. In other
words the same aperture and shutter combination. The use of the
basic auto-exposure mode is not recommended. It can lead to too
great a variance between images. Some, but by no means all
stitching software, has blending and exposure correcting features,
but there is still a limit to what these can do in correcting the
imbalance between adjacent images.
The best course of
action is to set the exposure manually, and use the same setting
across all shots taken. If there appears to be a wide variance in lighting levels, then the
brightest part of the scene to be captured should be used to set the
exposure, to ensure there are no blown highlights.
To
achieve a really good image many photographers try to use as small
an aperture as possible to ensure good depth of field, just as you
would with a landscape shot. Unless light levels are high this means
of course in most cases the use of some kind of support, in the form
of a monopod or tripod, is required.
Following the rules
So how do you follow
these rules? By taking care when shooting hand held, and using
equipment to help you where you can if you wish. We'll take a look at using
monopods, tripods and other equipment in minute but first let's take
a look at taking shots hand held.
Hand
held
The situation that most people will find themselves
in sooner or later, is wanting
to take some shots they can stitch together on the spur of the
moment. For it's more than likely that unless you are
very keen on producing stitch or panoramic images you won't necessarily go
out with the express intention of doing so. Even if you are going
into a city to take some shots it doesn't always follow that
everything you will want to capture is going to be too big to fit in the
frame. But it's often the case that things turn out to be much bigger than you thought, looking through a
camera's viewfinder, than they do viewed with the naked eye.
So
you think, I know, I'll take some images and stitch them together
later. The trouble is it's notoriously
difficult when you are using a camera hand-held to keep it aligned
between shots, either horizontally or vertically. It's also very
easy to twist the camera around the lens axis so that subsequent shots are aligned in one
plane but not the other. You only have to look at photos people often
take,
where the horizon ends up at an angle to the horizontal plane of the
image to see the difficulty that exists.
Ideally you're looking to take a series of
shots as in fig 2. But often, if you are not careful, you end up
with a set looking either like fig 3. or fig 4. or sometimes both,
fig 5. These show horizontal format images but the same applies
equally to those taken vertically, portrait format. If your taking
more than one row of shots then errors like those in
fig
3 can be overcome, the rows above or below filling in some of gaps.
But images skewed around the lens axis like those in figs 4
& 5 lead to image mis-alignments that are difficult to correct.
To
combat this the task is to
ensure that you stand correctly when taking shots and move around
the camera as in fig 6. This will ensure that the pivot point is
reasonably close to the ideal. Making yourself the pivot point and
swinging the camera around you in an arc will make things much
harder for the software, although this is the most natural action to
take, so you have to remind yourself not to do it. fig 7. You may
sometimes get away with this if taking just a couple of shots for a
Panoramic type stitch with the focused distance being at or
near infinity, otherwise you will end up frustrated and disappointed
at the lack of stitching success.
Just
pivoting the camera/lens around its centre point might not sound ideal after
what we have written regarding the nodal point of lenses, but we
have found that it seems to work reasonably well with series of
shots that do not cover a wide arc. We invariably take
hand held shots, even for multiple row stitching, and most of the
time it works out, although not always. Some of this
might be down to the software we use which we will discuss further
on, or just luck, although as with anything a bit of practice always helps. Going around locally and just taking practice stitch shot sequences
will soon show you what's possible and what's not, and where you might need to improve your technique to
succeed. It's also useful for
getting to know your software and what it's capable of. And of course
with digital it costs little, just a bit of time and effort.
Simple
Panorama's
To
take a sequence of shots you need to look through the viewfinder and
see how much of the scene is captured. Generally for a simple
panoramic two or three shots are involved. Keeping it to two is okay
but taking three with more overlap is better. Take one straight
ahead, then one to each side. Make a mental note of where the image
elements are in the first shot so you know just how far to pivot the
camera to make the subsequent shots overlap the first.
Here
is a very simple cylindrical panoramic stitch involving just two images. It's of
the Roman Amphitheatre in Verona, Italy. When we
took these shots of it in 2002, using a 3mp digicam, a famous
Italian film director was in the process of staging the play Anthony
and Cleopatra in it, which explains the props. Originally we
generated the stitch image using the Photomerge application
in Photoshop Elements, but this current stitch was created using the PTGui
stitching software which we now use. Verona is of course famous
as the setting for the Shakespeare play Romeo and Juliet.
Roman
Amphitheatre, Verona, Italy.
This
panorama is a classic example of when moving further back to
capture a scene just isn't possible. The digicam we were using had a
34mm [equiv] view at the wide angle end - the two images used being
shot at this focal length - but moving further back only meant that
other features in the square in Verona where the amphitheatre is
located, trees, fountains, etc, then blocked some of it from view.
There was no way we could capture it in one shot, and of course the
position of the large props just begged capture from the perspective
we had chosen.
The
ability to use a much wider angled lens might well of resulted in
the capture of the same scene in one shot, but the perspective
would have been different. As it is, the props, though large, do not
dominate, there's a fairly equal balance between them and the
amphitheatre. A wider view would have resulted in the props being
very dominant and un-balanced the scene as we wanted it, the
amphitheatre then becoming just scene setting background detail.
Multiple
Image Panorama's
Taking more
shots or multiple rows is just an extension of this basic concept, although we
have to say that taking really wide panoramic images involving many
shots in each row hand held is not really a good idea. To do that successfully
you really need to use some kind of camera support, and pay closer
attention to the lens nodal point.
However........
having said that, here's an image that is composed from 8
shots, 2 rows of four, all hand held. The task we had here was to capture the front
entrance to this shopping mall, which is situated down a pedestrian
side street. Once
again lack of space to capture in one shot was the problem, along
with the requirement to keep the final perspective as near to normal as
possible.
Again
this
is not from a flat stitch but a cylindrical one, as you
might see from the curved front of the extreme right hand side
building, which is flat in real life although the rest of the building is
curved as in the image. This is as a result of using this format as opposed to a
flat projection. Flat projections keep all straight lines straight
but produce distortions (compressing the image vertically/stretching
horizontally) at the edges of images that cover
greater than about 120°. A cylindrical stitch produces curved lines
but less distortion in other areas. It has the opposite effect of a
flat planar stitch compressing horizontally/stretching
vertically.
What type you choose very
much depends on the type of panorama being stitched,
buildings/landscape, and the perspective you
require or prefer. The best Panoramic software allows you to switch
views to see the results before rendering the finished stitch.
You must of course understand that there
is a difference between image distortion caused by optical problems
and perspective distortion caused by the width in degrees of the
stitch being assembled. In fact due to perspective distortion problems flat
planar images over about 100°- 120° are not really recommended.
Our software is designed not to stitch them and automatically sets
cylindrical stitching, the planar option being unavailable.
There's
another fault we have deliberately left in this stitch. The
woman in the green coat exiting the building. Two images of her exist. This is as a
result of taking sequential shots and merging them. Her position had
changed because she was moving between shots. And that's another
downside. You can't take action stitch shots, it just doesn't work.
Distant
Panorama's
Just in case you get the idea we don't take
distant panoramic
stitch shots
here's one of a well known building, the Sydney
Opera House. The use of a panoramic view here helps to put it into
perspective in relation to it's surroundings. Again this is a
cylindrical stitch using just
three shots taken hand held, this time from the middle of Sydney
Harbour bridge. The view covered is around 160° but doesn't look it
because of the distance and the large expanse of water in front
which helps to hide perspective distortions.
Sydney
Harbour and Opera House.
On
the next page we'll deal with the subject of taking shots using
equipment.
Monopods, Tripods, and the use of special Panoramic
Heads.