Binning
Binning (cf. the German term " Gebinde ") is used in several - mainly technical - contexts.
Digital photography
In digital image acquisition with optoelectronic sensors, such as in digital photography , what is known as binning is understood as the combination of neighboring picture elements ( pixels ) in a digital camera . The formation of pixel blocks results in a higher light sensitivity per virtual image point, with the signal-to-noise ratio being improved because the noise is statistically distributed. In return, however, the image resolution is also reduced in accordance with the number of combined pixels, i.e. H. the picture becomes coarser.
With binning, pixels are combined within a row and / or a column. This combination is done by adding the brightness values , either in an analog way by physical charge addition and charge transport on the image sensor itself (see CCD sensor ) to the local line amplifier or digitally by adding the digitized values. Analog binning increases the camera's frame rate because fewer pixels have to be read out and digitized. In an illustration of z. B. 4: 1, the pixels are output with half the horizontal frequency and the vertical frequency is maintained. Because of half the number of lines, an image takes half as long. The exposure time is halved per virtual image point, but this already has theoretically four times the amount of charge. Sometimes the neighboring pixels of a line are combined in analog form and the neighboring lines digitally.
Another advantage of the method is the reduction of the bandwidth during the transmission to the subsequent processing system, especially if very high-resolution sensors are used, the number of pixels of which is not required for the specific application - but where a high dynamic range is important and a later grouping anyway would be necessary. In the case of various industrial sensors, the grouping takes place not only through the chip matrix, but also through a digital signal processor connected downstream of the analog-digital converter . Such sensors can sometimes be configured externally via a digital interface, for example to set the desired image resolution and the pre-filtering. Some systems carry out a permanent brightness adjustment and adjust the optimal binning ratio to the exposure.
Light emitting diodes
In the production of light-emitting diodes, there are manufacturing-related small deviations in the color temperatures and brightness values that are noticeable in a direct comparison (for example when combining several light-emitting diodes in one luminaire). By combining equipping of devices, larger groups of the same intensity can be formed again by subgroups of light emitting diodes of known intensity in order to e.g. B. to achieve a balanced and even light intensity on large screen projection walls or displays with backlighting.
The division of the products into the various finely graded classes after production is called "binning", whereby accordingly finely graded parameters are used to sort the products into so-called bins , that is, the LEDs are a group of the same lighting intensity assigned. With white LEDs, the interior usually comprises four areas, the so-called “flux bin”, the threshold voltage, the light output and the color cast (unavoidable with LEDs). Colored LEDs are also offered with spectrally selected tolerances. The manufacturers provide information about the bin and the assigned properties in the data sheets of the LEDs.
Data processing and analysis
In the context of general data analysis, binning is understood to be class formation as a preprocessing technique. The target quantities of the attributes are divided according to size in increasing intervals - so-called bins (English for containers ). All attribute values are then replaced with the representative of the interval in which the value is located. This representation value, also known as the interval label , is often about the average or the median . It is therefore a form of quantization .
This form of binning can be used on the one hand to reduce the number of values of a given attribute and thus the amount of data. Furthermore, the consequences of small deviations in the attribute values, for example due to measurement errors, can be reduced.
Example of data reduction
| Time | Temperature in ° C | Time | Temperature in ° C | |
| 08:00 | 0.2 | 12:00 | 3.5 | |
| 09:00 | 0.7 | 13:00 | 4.9 | |
| 10:00 | 1.3 | 14:00 | 6.3 | |
| 11:00 | 2.1 | 15:00 | 8.1 |
When measuring the temperature profile on a day, the data are as shown in the table above. For the temperature range from 0.0 to 9.9 ° C, with one decimal place, the target amount consists of 100 values. This can be achieved with 7- bit words, which corresponds to a data volume of 56 bits with 8 data values.
| Time | Temperature interval in ° C | Time | Temperature interval in ° C | |
| 08:00 | [0.2) | 12:00 | [2.4) | |
| 09:00 | [0.2) | 13:00 | [4.6) | |
| 10:00 | [0.2) | 14:00 | [6.8) | |
| 11:00 | [2.4) | 15:00 | [8.10) |
If you use binning with an interval length of 2 ° C, i.e. with intervals [0.2) to [8.10), you get values as shown in the second table. The respective mean value of the interval can be used as a representative. Data processed in this way has a target set of 5 values, which can be implemented with 3-bit words, i.e. 24 bits with 8 values.
This has reduced the data size by 4 bits per value and the data volume from 56 to 24 bits. This means a trade-off between data size and information, since the values are then less precise.
Application in statistics
One application of binning is the graphic representation of the frequency distribution of a sample using a histogram . First, binning is carried out in order to divide the values into classes. Then the absolute or relative class frequency is chosen as the respective representative and shown in a bar chart.
Web links
- Wolfgang Paech: The term binning or on-chip binning. In: Glossary - Explanation of technical terms for CCD cameras. Baader Planetarium, 2005, accessed on April 8, 2009 (Binning in Astronomy).
- http://www.itwissen.info/definition/lexikon/Binning.html
Individual evidence
- ↑ Alexander Linder: Web mining - the Swarovski case study: theoretical foundations and practical application , Springer, 2005