by Jamie Beckett
When digital cameras were first introduced,
camera makers strove just to match the resolution
and image quality of traditional film cameras.
Now, thanks in part to research from HP Labs,
HP's newest digital cameras have capabilities
film photographers only dream about.
The Photosmart R707, introduced in May, inaugurated
a new line of cameras that are able to eliminate
red-eye, stitch together panorama photos in preview
mode and produce pictures closer to what our
eyes see -- all inside the camera, without a
"HP has been driving the pace of new technologies
in digital image science for years," says
Frank Carau, technology research manager for
HP Digital Cameras. "Our scanners, printers
and digital cameras show the results."
HP Labs began feeding the technology pipeline
for digital cameras as early as 1997.
"Initially, our research was focused on
getting the color right, improving sharpness
and contrast -- those were the big issues,"
says Dan Tretter, a researcher in the Imaging
Systems Lab. "On the hardware side, resolution
was the big issue."
Although mainstream digital cameras have not
yet exceeded the resolution (measured by apparent
sharpness and visible detail) of high-quality
film cameras, the quality of digital has improved
so much that the difference is indistinguishable
for the common 4 x 6 prints.
As a result, researchers shifted their focus
to building on digital’s strengths.
"Digital cameras can do things film cameras
can't," says Qian Lin, manager of
the Imaging Technology Department in HP Labs.
"They can capture a wider area than optical-lens
cameras are physically capable of. They can capture
high-resolution video with sound. In the future,
we think cameras will have more processing power
and be able to actually 'understand' what
For HP's new cameras, researchers in the
company's imaging business and in HP Labs
in Palo Alto, Calif.; Bristol, England and Haifa,
Israel together created a package of advancements
called Real Life technologies, all aimed at making
it easier to take crisp, real-to-life photos.
Driven by an image engine co-developed by scientists
at HP and Texas Instruments, Real Life technologies
encompass not only in-camera red-eye removal
and panorama preview, but also adaptive lighting
to bring faces out of the shadows and details
out of backgrounds in high-contrast photos, image
advice and an image noise filter.
Some of the technologies appearing for the first
time in the R707 include:
In-camera red-eye removal
Red-eye occurs when the eye's
retina reflects the camera's flash. It's
particularly a problem in digital cameras because
the flash and the lens are typically closer together
than in film cameras, which reflects more light
into the eye.
Most digital cameras have a pre-flash that reduces
red-eye, but it also requires subjects to remain
posed throughout a series of flashes, resulting in
a subject who, all too often, has moved out of the
pose or blinked and is photographed with their eyes
When researchers began work on red-eye
removal, their initial goal was to develop
a software tool that allows a user to
select a region containing a red-eye and
correct the red tint.
Since refined, the patented algorithm
now works automatically,
red areas in a photograph, then performing
tests to determine whether that red is
part of a human eye. If so, the algorithm
removes it and previews the changes on
the camera's LCD screen. Users may accept
or reject the changes before saving the
Walter S. Mossberg, the Wall Street Journal
and his assistant
the R707’s red-eye
for the newspaper.
"We came away positively
impressed," he wrote.
The eyes and brain naturally deal
with very high-contrast scenes -- those with
dark shadows or bright areas. HP's Adaptive Lighting
technology uses these same processing principles
to adjust previously difficult images such as
skiers in the snow or someone indoors standing
in front of a window.
"One of the most common photography problems
is back-lit photos," says Tretter. "When
you take a picture, you don't want to have
to worry if there's a window behind the person,
or what direction the sun is coming, or if person
is standing in shade."
Adaptive Lighting, developed by a team from
HP's imaging business and HP Labs, "re-lights"
photos at the moment they're shot to preserve
life-like contrasts. Trying to perform the same
task using imaging software is both difficult
and less effective because critical picture information
is lost when the image is compressed to a JPEG
and downloaded to a computer. Adaptive Lighting
processes all the data when re-lighting an image.
(Adaptive Lighting debuted in the HP Photosmart
945, the world's first camera to use Retinex
theory so photos look more like what our eyes
see. Four patents are pending on this technology.)
In-camera panoramic view
Today, digital photographers can take several photographs
in sequence and use PC software to stitch them together
into one panoramic shot. But post-production work takes
time, and it’s difficult to get good results.
often get home only to find that they've missed
an important part of the image, the images aren't
aligned properly or something else went wrong while
taking the photos. At that point, it's too late.
The in-camera panorama
features up to five
images in a series, automatically aligning and merging
the edges of each subsequent photo into one continuous
shot. Photographers can preview and scroll this panoramic
view on the camera's display screen to immediately
determine if something went wrong. If so, they have
a chance to re-shoot the scene. To make these factors
consistent across each photo within the series, the
camera automatically manages color, exposure and zoom.
(To see panoramic images or read more about the research
behind this, go here.)
HP adaptive demosaic technology
This accurately fills in full color information for every pixel, creating
crisp, sharp images with outstanding resolution. For digital still cameras,
color is usually captured by placing a red, green, or blue filter over
each pixel in a checkerboard pattern known as a mosaic.
Next, a "demosaic" algorithm fills in the
missing two-thirds of the color information. Obviously,
filling in this much information makes the quality
of the final photo entirely dependent upon the
Most digital cameras use a demosaic algorithm that averages nearby
measured pixel values along horizontal or vertical directions to estimate
missing color information. This method leads to excellent resolution
in vertical and horizontal directions but poor resolution along a
By contrast, HP's Adaptive Demosaic uses a decision equation to pick
one of 256 possible averages of the nearby measured
color values. This
set of averages covers all possible orientations
of image edge detail so that images
have excellent resolution in horizontal, vertical
and diagonal directions.
This analyzes photos for problems such as over or under exposure or
fuzzy focus and displays feedback on how to adjust
settings to create a better picture in future shots.
(Above: The photo on the left has a problem with
white balance. In the photo on the right, the problem
has been corrected.)
HP scientists continue their efforts to make HP's
cameras even smarter. For example, cameras already
have the ability to record what time a photo was
taken, by which camera, at what focal distance
and at what exposure -- information that helps
people search and retrieve images, says Department
"We want to go beyond that," she adds. The team is working
to expand the metadata included in photo files
to include where the picture was shot, which direction
the camera was pointing and other similar information.
They're also investigating face
detection so that
photo files will
be able to indicate
if there are faces
in the picture, and
if so, how many.
include a "translating
camera" that translates signs
as users photograph
them and technologies
that make it easy to add audio, video,
music and other media
to a photograph.
"It's all about making it
easier for people
to capture and share
and relive their