EMVA officially unveils the new release 4.0 of the EMVA 1288 Standard for objective characterization of industrial cameras. The release takes into account the rapid development of camera and image sensor technology.
Until the previous Release 3.1 dated back December 2016, the application of the EMVA 1288 standard with a simple linear model was limited to cameras with a linear response and without any pre-processing. While this model is being continued with some improvements in the ’Release 4.0 Linear’, a new module ‘Release 4.0 General’ has been added in the latest release. With it, the characterization of a non-linear camera or a camera with unknown pre-processing is possible even without any model due to the universal system-theoretical approach of the EMVA 1288 standard. Just as with the linear camera model, all application-related quality parameters can be measured in this way. With both modules “Linear” and “General” the same measurements are performed. Depending on the camera characteristics, the proper evaluation either according to the linear or general model is applied.
In addition, Release 4.0 includes numerous expansions to characterize the latest generation of image sensors and cameras according to the application. The most important of these are:
- Extended wavelength range from UV to SWIR range.
- Raw data of any given image acquisition modality can now be characterized according to the standard.
- The versatile and universal analysis tools of the EMVA 1288 standard can also be applied to quantities calculated and derived from multiple channels. For polarization image sensors, these are, for example, the degree of polarization and the polarization angle.
- Inhomogeneities are measured in detail and now decomposed into column, row, and pixel variations. They can now be determined with a new method at all intensity levels from just two captured images.
- Optionally, cameras with optics or with illumination as given by the position of the exit pupil of the optics for which the image sensor was designed can be measured according to the standard. Thus, the standard is now also suitable for image sensors with pixels shifted towards the edge.
- A more suitable measure for the linearity of the characteristic curve is introduced.
I personally don't like the new definition of linearity. Just report the mean value will hide information. You can find sensors with pretty high linearity in a certain small range, but mean linearity of whole range is still very small. Nobody would like to use these sensors. I prefer to report max and min values in the whole range.
ReplyDeleteI too dislike their linearity characterization approach. For example, their recommendation is to measure linearity in 5%-95% full scale range. That ignores a low-light non-linearity coming from the image lag. In fact, image lag is not mentioned in the doc at all, even though it might be a major factor in low light performance evaluation, depending on a sensor model.
DeleteThere is s much data taken to make a PTC that linearity should just be reported as a graph of the first derivative. If a couple of numbers are needed for a table, report, for example, the band from 5-95%, 0-50% or some other commonly-used band.
ReplyDeleteThe same data could be reported for HDR images after any reconstruction needed.