White balance

The gray world assumption as well as the white patch retinex algorithm are used frequently for automatic white balance. The popular draw utility written by Coffin (2004) scales each color channel using the average as an automatic white-balance option. After rescaling, the white point is set at the 99th percentile. In other words, all channels are scaled equally such that only the top 1% of all pixels are clipped. This assumes that there are only a few highlights. 

VICTORIA And white balance, I didn t know what that was— KARIN I was trying to give her the crash course, some basics, and then when she was down there we would talk. And I d ask, "What d you guys get " and she d tell me, and I d [surest], "What about this and this " She s really good at interviews and knowing how to get the story, and I was more about the technical "Did you get shots of this," and b-roll, you know—... 

An important feature of the color perception is that, besides the raw sensing of tiie color world, the brain processes the information. One of these signal processings deals with the chromatic adaptation to what the brain perceives as the more intense and pleasant white. This is similar to what is known in digital photography as the white balance. This aspect is important as soon as we compare colors together, because a color coordinate is always defined relative to a reference stimulus that is processed as the white color. The reference white depends on the color space that is used to measure the color. 

The flow chart of the color acquisition procedure is given in Figure 3. The color machine vision system should be white balanced before taking measurement of samples. This can be accomplished by first measuring a white reference target to establish a base line. The R,G,B values of a sample are then acquired. To compare the machine vision measured tri-stimulus values against that of published standard color plates, R,G,B values from CMVS were transformed to L a b values. This transformation required two steps first, R,G,B is converted to X,Y,Z by Eq. 3. Next, Eq. 4 is used to convert X,Y,Z to L, a, b color coordinates. 

White balancing is another essential step in achieving accurate color measurement and it can be done by adjusting the sensor s output and the digitizer s... 

In addition to sensor QE, overall camera response of the different color channels depend upon several additional factors (1) the scene iUuminant spectral power density (2) the camera lens transmittance and (3) the infrared cutoff filter transmittance. The combined response from contributions (l)-(3) and the sensor QE is shown for a representative camera in Fig. 8. The assumed illuminant is CIE D65, with the spectral density converted from watts (per unit area per unit wavelength) to photons, so that the units cancel properly with those of the QE. The area under the green curve is highest, implying greatest sensitivity in this channel in daylight. Automatic white balance, discussed in the next section, adjusts gains to compensate for the lower areas under the red and blue curves, to yield an approximately neutral reproduction of the scene. 

Figure 9 shows some of the operations in a typical image processing pipeline. There are some variations possible in the ordering of the operations, and a few can be done in either the analog or digital domains, or both (e.g., black level subtraction and white balance). 

Black level subtraction Bad pixel correction White balance Demosaic ... 

---