For more than 20 years, various solutions have been found to the problem of HDR
imaging. It is only more recently that significant progress has been made in the field
of rolling-shutter and global-shutter CMOS image sensors, which provide a direct
HDR image output without software processing and with good image quality.
Several XDR sensor types have been discussed that can be useful in many
HDR applications, usually where a low-processing application or a fast solution
is necessary, or when the scene is moving relative to the camera. Each sensor
type has advantages and disadvantages. It has been shown how the performance of
XDR sensors can be measured or calculated using SNR, dynamic range factor, and
information, and how sensors can be compared to select a sensor for an application.
The most useful types of sensors today are multiple-segment response sensors and
those using advanced logarithmic pixels. For higher performance, such sensors
usually require larger pixels; this is not compatible with the ongoing race for higher
Software methods were then discussed (which are useful in photography of still
scenes), and one special example of a software-based high-speed machine vision
application was given. Some sensors provide a simplified version of this software
method for real-time operation and a small increase in dynamic range. Dedicated
hardware platforms provide limited real-time solutions. It was also seen that optical
path should be designed carefully, especially regarding optical veiling glare and
Now that sensors are capable of 150 dB or more using various design techniques,
and that software methods exist to go beyond 180 dB, it is time to work on the
improvement of other components of these systems such as control algorithms,
optical path, and testing and characterization processes.
Finally, file formats compatible with HDR images were briefly discussed, along
with testing methods for HDR cameras and systems.