Vision Research Inc has recently adopted EMVA1288 to measure and specify the performance of their Phantom high-speed cameras. For too long the high-speed camera industry has published specifications using standards that do not apply well to digital CMOS based sensors.
For many years the high-speed camera industry has adopted the old photographic film ISO standard and variations of it. ISO makes it difficult as a distributor or end user to compare digital CMOS high speed cameras from different manufacturers without testing them. It has always required a shootout which is often inconvenient and slow.
ISO is a standard that suited film-based imaging, but it is poor at characterising high-speed digital cameras because it misses some of the critical specifications required to truly evaluate high speed cameras.
ISO when used with digital CMOS high speed cameras is largely qualitative and inconsistent in its measurement of sensitivity, a critical factor of the performance of a high-speed camera. Often users interested in comparing cameras before purchasing, put them side by side to see which produces a brighter image in order to determine sensitivity. What they can fail to identify is the noise and loss of dynamic range in some cameras from manufacturers that choose to hide noise at the expense of true performance i.e. they create an artificially nice-looking image but compromise other performance parameters in doing so. This practice is not scientifically accurate and leads to false conclusions in analysis and measurement of images.
Developed by the EMVA the EMVA1288 standard is a camera standard that characterises digital CMOS camera performance in a scientifically robust way. Standardised equipment is supplied for testing cameras and to generate a consistently comparable report. The standard also defines the test method and so all facets of generating performance specifications is consistent and reliable regardless of the camera and its manufacturer.
The EMVA1288 standard treats a camera as a black-box model with light measured as an input and digital number measured as an output with both average and variance values of each being measured. The black box design is simple and sophisticated and models the camera with;
• A Quantum Efficiency
• A single source of Noise called Temporal Dark Noise or Read Noise (which is in fact a combination of several noise sources)
• A Gain
• A Quantisation (including error) to convert to Digital Numbers DN.
By measuring the inputs and outputs of cameras in the EMVA1288 test setup and using the Poisson relationship inherent in the photon shot noise behaviour of light, where the average value is equal to the variance, it is possible to determine the overall Gain factor in the camera. From there it is a matter of mathematics to accurately determine the other parameters that clearly and scientifically characterise a camera.
|