Graphics Reference
In-Depth Information
exposures in the middle section of the response curve is known as the
linear range
or
working range
of the device. Ideally the recorded value is directly proportional
to the exposure. In practice however, there is some variation. Outside the linear
range, the pixels cannot be expected to faithfully represent scene radiance values.
The dynamic range is therefore limited to the middle range in practice.
Under normal shooting conditions, a photographer tries to set the exposure
time (the shutter speed) and the lens aperture so that the exposure values corre-
sponding to the scene radiance values fall within the linear range of the camera.
But in many scenes, particularly those in direct sunlight, the range of values in
the scene exceeds the dynamic range of the camera. Photographers often divide
the apparent brightness in a scene into three conceptual ranges:
shadows
,
mid-
tones (photographic brightness range)
,and
highlights
. The camera is normally
set to capture the mid-tones, and detail is expected to be lost in the shadows and
highlights. One approach to faithfully capturing a greater dynamic range is to
take a collection of photographs of increasing exposure time, a process known as
bracketing
the exposure. A collection of bracketed images can be digitally com-
bined into a single image of a higher dynamic range, but doing this accurately
requires precise knowledge of the camera response curve.
6.1.2 HDR Response Curve
The paper “Recovering High Dynamic Range Radiance Maps from Photographs”
by Debevec and Jitendra Malik provides a simple, effective method for construct-
ing a high dynamic range image from a collection of photographs [Debevec and
Malik 97]. The motivation for the work came fromDebevec's short film
The Cam-
panile Movie
, described in Chapter 5. Debevec had originally intended to shoot
the movie on a clear day, but the clock faces in direct sunlight were so much
brighter than the clock faces in shadow that the photographs could not simulta-
neously capture the appearance of both. Setting the exposure time to capture the
sunlit faces caused the shadowed faces to come out too dark; using a long enough
exposure time to capture the shadowed faces made the sunlit faces appear “blown
out.” He found an easy way of getting around this problem: he simply shot
The
Campanile Movie
on a cloudy day. But his interest in high dynamic range imaging
was piqued.
Extending the effective dynamic range of photographs and prints is an old
problem in photography. By the time Debevec became interested in the problem,
a number of methods to create a linear response curve from of wide range of
brightnesses existed in the computer vision literature. Debevec was particularly
interested in a method for recovering a linear response from digital camera images
proposed in a 1995 paper by Steve Mann and Rosalind W. Picard [Mann and
